tutorial video tutorial video Discover the CiteMap in 3 minutes
Source(s) of the information:
Source(s) of the information:

Lawyers, other representatives, expert(s), tribunal’s secretary

Recourse to article 22.6 of the DSU by the United States - Decision by the Arbitrator

ABBREVIATIONS

AbbreviationDescription
2002 Farm Bill Farm Security and Rural Investment Act of 2002, Public Law No. 107‑171, Section 10816, 116 Stat. 134
2008 Farm Bill Food, Conservation, and Energy Act of 2008, Public Law No. 110‑234, Section 11002, 122 Stat. 923
2009 Final Rule Final Rule on Mandatory Country of Origin Labeling of Beef, Pork, Lamb, Chicken, Goat Meat, Wild and Farm-Raised Fish and Shellfish, Perishable Agricultural Commodities, Peanuts, Pecans, Ginseng, and Macadamia Nuts, United States Federal Register, Vol. 74, No. 10 (15 January 2009), pp. 2658‑2707
2013 Final Rule Final Rule on Mandatory Country of Origin Labeling of Beef, Pork, Lamb, Chicken, Goat Meat, Wild and Farm-Raised Fish and Shellfish, Perishable Agricultural Commodities, Peanuts, Pecans, Ginseng, and Macadamia Nuts, United States Federal Register, Vol. 78, No. 101 (24 May 2013), pp. 31367‑31385
amended COOL measure COOL statute together with the 2009 Final Rule, as amended by the 2013 Final Rule
AMS Agricultural Marketing Service of the USDA
BCI business confidential information
compliance panel panel in the Article 21.5 compliance proceedings in US – COOL
COOL country of origin labelling
COOL statute Agricultural Marketing Act of 1946, 60 Stat. 1087, United States Code, Title 7, Section 1621 et seq., as amended by the 2002 Farm Bill and the 2008 Farm Bill
DSB Dispute Settlement Body
DSU Understanding on Rules and Procedures Governing the Settlement of Disputes
GATS General Agreement on Trade in Services
GATT 1994 General Agreement on Tariffs and Trade 1994
original COOL measure COOL statute together with the 2009 Final Rule
original panel panel in the original proceedings in US – COOL
TBT Agreement Agreement on Technical Barriers to Trade
USDA United States Department of Agriculture
Working Procedures Working Procedures of the Arbitrator
WTO World Trade Organization
WTO Agreement Marrakesh Agreement Establishing the World Trade Organization

1 INTRODUCTION

1.1 INITIAL PROCEEDINGS

1.1.
The present arbitration proceedings arise in the disputes initiated by Canada and Mexico concerning the United States' country of origin labelling (COOL) requirements for meat products.
1.2.
On 23 July 2012, the DSB adopted the original Appellate Body reports in these disputes, and the reports of the original panel as modified by the Appellate Body.1 The findings adopted by the DSB were that the COOL measure at issue in the original proceedings (the original COOL measure2) was inconsistent with Article 2.1 of the TBT Agreement because it accorded less favourable treatment to imported livestock than to like domestic livestock.3
1.3.
On 4 December 2012, following referral to arbitration under Article 21.3(c) of the DSU, an arbitrator determined that the reasonable period of time for the United States to implement the DSB's recommendations and rulings would expire on 23 May 2013.4 At the DSB meeting on 24 May 2013, the United States announced that, in order to come into compliance with the DSB's recommendations and rulings, the United States "had issued a final rule that made certain changes to the country-of-origin (COOL) labelling requirements", and that these actions "brought the United States into compliance" with those recommendations and rulings.5
1.4.
On 19 August 2013, Canada and Mexico requested the establishment of a panel under Article 21.5 of the DSU, to determine whether the "amended COOL measure"6 brought the United States into compliance.7 On 29 May 2015, the DSB adopted the Article 21.5 Appellate Body reports in these disputes, and the reports of the compliance panel as modified by the Appellate Body.8 The findings adopted by the DSB were that the amended COOL measure violated Article 2.1 of the TBT Agreement and Article III:4 of the GATT 1994 because it continued to accord less favourable treatment to imported livestock than to like domestic livestock.9
1.5.
On 4 June 2015, Canada filed a request with the DSB for authorization to suspend concessions or other obligations under Article 22.2 of the DSU.10 In its request, Canada sought authorization to suspend concessions and related obligations in the goods sector under the GATT 1994 to an annual value of CAD 3,068 billion.11
1.6.
On 4 June 2015, Mexico filed a request for authorization to suspend concessions or other obligations under Article 22.2 of the DSU.12 In this initial request, Mexico sought authorization to suspend concessions and related obligations in the goods sector under the GATT 1994 in an annual amount of USD 653.5 million.13 On 12 June 2015, Mexico filed a corrigendum, correcting the amount to USD 713 million.14 On 17 June 2015, Mexico re-submitted the request for authorization to suspend concessions or other obligations in the amount of USD 713 million.15

1.2 REQUEST FOR ARBITRATION AND ARBITRATION PROCEEDINGS

1.7.
On 16 June 2015, the United States communicated to the DSB its objection to Canada's proposed level of suspension of concessions or other obligations.16 At its meeting of 17 June 2015, the DSB took note that the matter raised by the United States had been referred to arbitration, as required by Article 22.6 of the DSU.17
1.8.
On 22 June 2015, the United States communicated to the DSB its objection to Mexico's proposed level of suspension of concessions or other obligations.18 As noted in document WT/DS386/37, the parties agreed that the matter had been referred to arbitration under Article 22.6 of the DSU.19
1.9.
The arbitration was undertaken by the original panelists, namely:

Chairperson: Mr Christian Häberli

Members: Mr Manzoor Ahmad

Mr João Magalhães

1.10.
A joint organizational meeting was held on 3 July 2015 to discuss procedural aspects of the proceedings. At this meeting, all parties agreed that the proceedings with respect to Mexico and Canada should be conducted together rather than separately. Furthermore, as discussed below in section 2.2, Mexico and Canada requested third-party status in order to be able to fully participate in each other's arbitration; the United States, while raising systemic concerns in respect of third-party rights in Article 22.6 arbitration proceedings, stated its support for full participation of Canada and Mexico in each other's arbitration.
1.11.
Additionally, the United States and Canada requested that the substantive meeting be conducted as an open hearing, by which public viewing of the hearing would be permitted. Mexico raised no objection to holding an open hearing.
1.12.
Finally, the United States and Canada requested that working procedures be adopted for the protection of Business Confidential Information (BCI), and Mexico agreed to the inclusion of BCI working procedures.
1.13.
Taking these considerations into account, and in order to accommodate the interconnected nature of the respective disputes, the Arbitrator: (a) adopted a harmonized timetable; (b) decided to hold a joint substantive meeting with the parties; (c) granted Mexico and Canada certain rights to participate in each other's proceedings, as further discussed below in section 2.2; and (d) decided to include the two decisions in one single document with the final sections containing the Conclusion and Decision being printed on separate pages with the appropriate document symbol relevant for each dispute. Furthermore, the Arbitrator granted the parties' request for an open hearing as well as the request to protect BCI. The Arbitrator accordingly adopted Working Procedures of the Arbitrator, BCI Working Procedures, Procedures for an Open Substantive Meeting of the Arbitrator, and the finalized timetable, and communicated those documents to the parties on 6 July 2015.
1.14.
In accordance with the timetable and Working Procedures adopted by the Arbitrator, on 10 July 2015, Canada and Mexico presented communications concerning the methodology for calculating the proposed level of suspension (methodology papers). Due to a corrigendum submitted by Canada one working day later in connection with its methodology paper, the United States was granted one additional working day to file its written submission in DS384. The United States provided its written submissions on 29 July and 30 July 2015, with regard to Mexico's and Canada's methodology papers respectively. Canada and Mexico provided written submissions to the Arbitrator on 12 August 2015. The Arbitrator sent written questions to the parties on 21 August 2015, to which the parties provided written responses on 1 September 2015. The Arbitrator held its substantive meeting with the parties on 15 and 16 September 2015. The parties provided written responses to an additional set of questions from the Arbitrator on 1 October 2015, and submitted comments on each other's responses to those questions on 8 October 2015.
1.15.
This Decision is structured as follows: we first address two procedural issues in section 2. We then turn to the substance of the proceedings and start by providing a brief overview of the parties' main arguments in section 3. Following this, we set out a number of preliminary issues in section 4 before undertaking the assessment of the proposed levels of suspension in section 5. Section 6 sets out our own determination of the level of nullification or impairment. Our conclusion and decision on the level of suspension of concessions or other obligations is contained in section 7.

2 PROCEDURAL ISSUES

2.1 WHETHER THE OBJECTION TO MEXICO'S REQUEST WAS PROPERLY REFERRED TO ARBITRATION

2.1.
As noted above, on 4 June 2015, Mexico filed a request for authorization to suspend concessions or other obligations under Article 22.2 of the DSU.20 In that request, Mexico sought authorization to suspend concessions and other related obligations in the goods sector under the GATT 1994 in an annual amount of USD 653.5 million.21 It also requested that a special meeting of the DSB be held on 17 June 2015 to consider its request. On 12 June 2015, Mexico submitted a corrigendum, circulated on 15 June 2015, correcting the amount to USD 713 million annually.22
2.2.
At the outset of the DSB meeting held on 17 June 2015, prior to the adoption of the agenda, Mexico asked that the item related to its request under Article 22.2 be removed from the agenda of that meeting in light of the corrigendum it had filed on 12 June 2015 and in order that the 10-day advance notice for circulation of documents be preserved.23 On 17 June 2015, Mexico re-submitted the request for authorization to suspend concessions or other obligations in the amount of USD 713 million annually, requesting that a special meeting of the DSB be held on 29 June 2015 to consider its request.24 On 22 June 2015, the United States notified to the DSB its objection to Mexico's proposed level of suspension and stated that "[a]ccordingly … the matter has been referred to arbitration".25 Thereafter, Mexico cancelled its request for a DSB meeting.26 On 26 June 2015, the Secretariat circulated a note indicating that "the parties agree that the matter has been referred to arbitration under Article 22.6 of the DSU", and noting the composition of the Arbitrator.27
2.3.
On 9 July 2015, the European Union communicated to the DSB its views regarding the communication from the United States circulated on 23 June 2015 "concerning certain recent procedural developments", and notably the reference in the United States' communication that Mexico's Article 22.2 request to the DSB "has been referred to arbitration, even though the DSB meeting originally scheduled to make the referral on 29 June 2015 was cancelled".28 In that communication, and subsequently in the DSB meeting on 20 July 2015, the European Union stated that it "does not agree that an Article 22.6 DSU request to the DSB may be referred to arbitration other than by the DSB."
2.4.
According to the European Union, the phrase "shall be referred" in Article 22.6 means that "there is an actor that does the referring and that actor is the DSB".29 In other words, it is the DSB that refers the matter to arbitration and the matter is not referred automatically when a notice of objection to a proposed level of suspension is filed.
2.5.
The European Union considered that the use of similar language in other provisions in the DSU, such as "shall be established" in Article 6 and "shall be adopted" in Articles 16.4 and 17.4, support its position, because in those cases the DSB is the actor that carries out those functions. The European Union also drew attention to the multiple references to the DSB in Article 22.6 (including that it is the DSB that grants authorization to suspend concessions). For the European Union, "this context strongly supports the view that it is also the DSB that refers the matter to arbitration."30 Finally, the European Union raised a number of "good reasons" for its view that the DSB must refer matters to arbitration, arguing that this view ensures that: (1) authority for binding dispute settlement "flows from the Members acting together, through the DSB"; (2) Members are informed in a timely manner of the scope and nature of the arbitration; (3) Members have an opportunity to express their views on the arbitration; and (4) Members have an opportunity to consider whether to seek to participate in the proceedings.31
2.6.
The European Union is not a party to these proceedings, and no party to the arbitration has raised any objection in respect of the referral of this matter to arbitration. Indeed, as noted above, the parties agree that the matter has been referred to arbitration. Nevertheless, there are instances in which an adjudicator remains under a duty to investigate issues that are not raised by parties to the dispute, particularly regarding issues of a fundamental nature related to its authority to preside over the proceedings. The Appellate Body explained in Mexico – Corn Syrup (Article 21.5) that:

[A] panel comes under a duty to address issues in at least two instances. First, as a matter of due process, and the proper exercise of the judicial function, panels are required to address issues that are put before them by the parties to a dispute. Second, panels have to address and dispose of certain issues of a fundamental nature, even if the parties to the dispute remain silent on those issues. In this regard, we have previously observed that '[t]he vesting of jurisdiction in a panel is a fundamental prerequisite for lawful panel proceedings.' For this reason, panels cannot simply ignore issues which go to the root of their jurisdiction – that is, to their authority to deal with and dispose of matters. Rather, panels must deal with such issues – if necessary, on their own motion – in order to satisfy themselves that they have authority to proceed.32

2.7.
Thus, there is a legal duty on panels to seize themselves of questions that are of a "fundamental nature", including the vesting of jurisdiction. We believe that this duty also applies to arbitrators. We recall in this context that in US – Section 110(5) Copyright Act (Article 25), the arbitrator considered that the principle by which an international tribunal is entitled to consider its own jurisdiction applies equally to arbitration bodies as it does to panels, and thus proceeded to examine on its own motion the question whether it had "the necessary jurisdiction".33
2.8.
On 15 July 2015, the Arbitrator communicated to the parties – namely Mexico and the United States – that it was considering whether any issues of a fundamental nature were present, particularly those that may go to the root of the Arbitrator's jurisdiction, in the context of the arbitration in DS386. The Arbitrator invited the parties to that dispute to provide their views on the issue.34
2.9.
In response to the Arbitrator's invitation, Mexico and the United States submitted a joint communication on behalf of both parties to the dispute.35 In the joint communication, Mexico and the United States stated that they did not see any fundamental issues that would require the Arbitrator to take action, and emphasized that both Mexico and the United States agreed that "the matter at issue was referred to arbitration by virtue of the filing by the United States of its objection to Mexico's request."36 Mexico and the United States noted that Members were fully informed about the arbitration through the request for authorization to suspend concessions37, the United States' objection to that request38, and the notification of the constitution of the Arbitrator.39 Mexico and the United States raised various considerations regarding the interpretation of Article 22.6, including: previous occasions in which matters had been referred to arbitration under Article 22.6 without any DSB action40; the text of Article 22.641; the text of Articles 6.1, 16.4, 17.14, and 22.7 of the DSU, which refer to "shall be"42; the applicable positive and negative decision-making rules under the DSU43; the authority and functions of the DSB44; procedures and provisions with respect to other arbitrations provided for under the DSU45; and procedural and timing implications.46
2.10.
Turning to our assessment of the issue, we begin with the text of Article 22.6, which states in relevant part:

When the situation described in paragraph 2 occurs, the DSB, upon request, shall grant authorization to suspend concessions or other obligations within 30 days of the expiry of the reasonable period of time unless the DSB decides by consensus to reject the request. However, if the Member concerned objects to the level of suspension proposed … the matter shall be referred to arbitration.

2.11.
The question whether the DSB must take specific action when an objection to a proposed level of suspension is notified in order to effect a referral to arbitration, or whether the objection itself has this effect, is a contentious issue among Members.47 We note that Article 22.6 provides in mandatory language that the matter "shall be referred" if the Member concerned objects to the level of suspension proposed. However, in using passive language without identification of the actor, Article 22.6 does not provide clear guidance on how this occurs. As noted by the parties and the European Union, similar passive language for actions that "shall be" carried out is used throughout the DSU, and the actor to whom such language refers differs based on the terms and context of the provision in question. For instance, the provision in Article 6 of the DSU that panels "shall be established" explicitly provides for this to be done at a DSB meeting, further stipulating that the DSB may decide by consensus not to establish a panel. No such explicit language is evident in the second sentence of Article 22.6 with respect to the referral of arbitration.
2.12.
Although the terms of Article 22.6 do not prescribe the manner of referral, there are contextual indications within the DSU suggesting that referral to arbitration need not be performed by the DSB. For example, a number of provisions of the DSU explicitly provide for arbitration proceedings in contrast to panel proceedings. "Arbitration" is contemplated under Article 21.3(c), Article 25, and Article 22.6. In arbitrations under Article 21.3(c) and Article 25, there is no explicit requirement of any action by the DSB to initiate the arbitration. Rather, Article 21.3(c) provides that the reasonable period of time for compliance "shall be … a period of time determined through binding arbitration", without further specification of the procedure or forum through which such arbitration is initiated. With respect to arbitration under Article 25, the DSU provides that "resort to arbitration shall be subject to mutual agreement of the parties" and that "[a]greements to resort to arbitration shall be notified to all Members sufficiently in advance of the actual commencement of the arbitration process", without explicit requirement of any action on the part of the DSB. Thus, these arbitration procedures under the DSU can be contrasted with the explicit requirements for the establishment of a panel described in Article 6, namely the initial request(s) by a Member and the subsequent establishment of a panel at a DSB meeting.
2.13.
The difference in explicit procedural requirements, as well as the difference in designation between "arbitration" and "panel", is consistent with Article 2 of the DSU, which sets out the functions and authority of the DSB. In particular, although the DSB has "the authority to establish panels", Article 2 makes no specific reference to the role of the DSB in relation to arbitrations. Further, it does not necessarily follow from its authority "to administer these rules and procedures" or other general functions that the DSB must carry out the specific act of referral to arbitration under Article 22.6, or under Articles 21.3(c) and 25.
2.14.
Further, we find it difficult to equate the arbitration referral procedure under Article 22.6 with that of panel establishment under Article 6 in light of the decision-making rule in Article 2.4, which states that "[w]here the rules and procedures of this Understanding provide for the DSB to take a decision, it shall do so by consensus." The establishment of panels authorized under Article 2.1 is based on negative consensus, as stipulated in Article 6.1. Similarly, adoption of panel and Appellate Body reports under Articles 16.4 and 17.14, respectively, is achieved through negative consensus decisions by the DSB, as is the authorization of suspension of concessions under Articles 22.6 and 22.7. Interpreting Article 22.6 to include a requirement of referral by the DSB implicates the decision-making rule that would apply to such action48, yet there is no explicit reference to such a decision in the text of Article 22.6.49
2.15.
We note that the initiation of dispute settlement proceedings without DSB action is envisaged in other contexts in the DSU, most notably for appeal procedures, which are triggered by notification of an appeal to the DSB pursuant to Article 16.4. In such circumstances, the DSB does not take any action to refer the matter to the Appellate Body, or indeed any action whatsoever in respect of the appeal, until the adoption of the reports. Other procedures in the dispute settlement process may also occur without DSB involvement, such as the suspension of panel proceedings and automatic lapse of the panel's authority under Article 12.12, which is triggered by the request of the complaining party. Based on such considerations, we are not persuaded that the initiation of every dispute settlement proceeding under the DSU, including arbitrations, must require action on the part of the DSB.
2.16.
At the same time, our approach does not diminish the exclusive role of the DSB in receiving and authorizing requests for suspension of concessions under Article 22, which applies irrespective of whether there is arbitration under Article 22.6. We also observe that, neither the parties nor any other Member, including the European Union50, have asserted any prejudice to its interests or rights under the DSU as a result of the manner of referral to this arbitration.
2.17.
As indicated above, the text of Article 22.6 does not explicitly require referral to arbitration by the DSB. Furthermore, the context found in other provisions of the DSU, particularly regarding other arbitration procedures, suspension and lapsing of panels, and initiation of appeals, suggests that it is not necessary for the DSB to have an active role in all dispute settlement procedures for them to occur. While agreeing that a resolution of this issue by Members would be desirable, the Arbitrator sees no reason in the present case to read such a formal requirement into Article 22.6.
2.18.
Therefore, the Arbitrator concludes that the procedural absence of formal DSB action in this case does not call into question the vesting of jurisdiction or the capacity of the Arbitrator to proceed. Hence, there was no reason in the present circumstances to suspend or terminate the proceedings on the basis of the manner of referral to arbitration.

2.2 THIRD-PARTY RIGHTS

2.19.
At the organizational meeting on 3 July 2015, Canada and Mexico requested to be third parties in their respective proceedings. Canada and Mexico clarified that they were seeking to have the right to be present at the entirety of the hearing and to have access to all written submissions. When asked specifically whether Canada was seeking a right to comment on issues not pertaining to its own case, Canada stated that it envisaged a right to comment where "issues of comparison" would arise.51 The United States raised systemic concerns in respect of third-party rights, taking the view that such rights were not provided for in arbitration proceedings. However, the United States supported "full participation" of Canada and Mexico in each other's case.52
2.20.
As noted in previous arbitrations under Article 22.6 of the DSU, arbitrators, like panels, have "a margin of discretion to deal, always in accordance with due process, with specific situations that may arise in a particular case and that are not expressly regulated."53 The DSU does not contain a specific provision on third-party rights in Article 22.6 arbitration proceedings, nor does it deny any such rights. Noting the absence of any such provision, previous arbitrators have denied requests for third-party status on the grounds that the party making the request could not show that its rights would be adversely affected through their inability to participate in the proceedings.54 However, arbitrators have authorized participation by Members not directly involved in the arbitration in certain situations. We note that in the two parallel arbitration proceedings in the EC – Hormones dispute, participation rights were granted because it was considered that the rights of the requesting Members "may be affected in both arbitration proceedings".55 In particular, it was noted that the product scope and relevant trade barriers were the same in both proceedings and that both arbitrators (composed of the same three individuals) might adopt the same or very similar methodologies.56 On these grounds, combined with the absence of any prejudice to the interests or due process rights of the respondent, the Members requesting suspension of concessions in the parallel cases were allowed "to attend both arbitration hearings, to make a statement at the end of each hearing and to receive a copy of the written submissions made in both proceedings."57
2.21.
In considering the requests of Canada and Mexico, we have taken into account the discretion of the Arbitrator to address procedural issues that are not specifically regulated in the DSU. Moreover, we consider the present circumstances to be similar to those present in the EC – Hormones arbitrations in respect of: similar products and relevant trade barriers being at issue; the potential need to adopt the same or similar methodologies in each proceeding; and, as confirmed by its agreement to participatory rights at the organizational meeting, the absence of any prejudice to the interests or due process rights of the United States. In particular, in view of the potential implications of adopting different methodologies, we cannot exclude the possibility that the rights of Canada and Mexico would be adversely affected through their inability to make a statement and submissions, and answer questions, in each other’s proceedings.
2.22.
In light of the foregoing, and based on the views of the parties expressed at the organizational meeting, the Arbitrator inserted the following paragraph in the respective Working Procedures of DS384 and DS386:

For the purposes of joining these proceedings with those in the parallel dispute [DS384][DS386], [Canada][Mexico] will be included in all communications of the Arbitrator and of the parties, including their submissions. [Canada][Mexico] will also be allowed to be present throughout the joint substantive meeting in DS384 and DS386.

2.23.
We have granted the above rights on the basis of our margin of discretion as described above. We note that these rights are not the same as those accorded to third parties in panel proceedings pursuant to Article 10 of the DSU. In particular, third parties in panel proceedings may make submissions in another party's case, including on issues not pertaining to its own case. Further, Canada and Mexico have been granted full access to all submissions and communications in each other's arbitration, including those made after the meeting with the Arbitrator.
2.24.
We consider that this affords Canada and Mexico the access and participatory rights requested, including the opportunity to comment on "issues of comparison" for purposes of each respective arbitration. Thus, Canada and Mexico have been allowed to fully participate in each other's proceeding to the extent necessary under the circumstances of these parallel arbitration proceedings.

3 MAIN ARGUMENTS OF THE PARTIES

3.1.
The parties have summarized their arguments in their executive summaries provided to the Arbitrator.58 In this section, we briefly set out the main elements of the parties' submissions made in these proceedings. We discuss in greater detail the parties' individual arguments in our analysis in sections 5 and 6 below.

3.1 OVERVIEW OF THE NULLIFICATION OR IMPAIRMENT CLAIMED BY CANADA AND MEXICO

3.2.
Canada and Mexico both claim nullification or impairment suffered as a result of the COOL measure59 in respect of two types of losses, namely (a) export revenue losses and (b) revenue losses from domestic price suppression. Both describe export revenue losses as a combination of suppressed prices and reduced quantities of livestock exported to the United States.60 In respect of losses suffered from domestic price suppression, both Canada and Mexico submit that due to "arbitrage" conditions in the North American livestock market, reduced export prices lead to suppression of domestic prices in their respective markets.61
3.3.
Canada claims losses in respect of four different categories of livestock, namely feeder cattle, fed cattle, feeder pigs, and fed hogs.62 Canada describes the level of its export revenue losses as totalling CAD 2,045 million63 and its losses from domestic price suppression as totalling CAD 1,023.1 million.64
3.4.
Mexico claims losses in respect of only one category of livestock, namely feeder cattle.65 Mexico submits USD 514.8 million in export revenue losses and USD 198.6 million in losses from domestic price suppression.66
3.5.
The respective methodologies used by Canada and Mexico to calculate the level of nullification or impairment overlap to a considerable extent. In particular, both use the same basic framework, namely a comparison between actual revenues obtained in a given baseline year after the expiry of the reasonable period of time (RPT), and estimated revenues that would have been obtained in that year absent the COOL measure.67 The methodologies of both Canada and Mexico use a one-year reference period following the expiration of the RPT pursuant to Articles 21.3 and 22.2 of the DSU.68
3.6.
In respect of the price estimation, Canada and Mexico follow the same general approach of econometrically estimating the COOL impact by means of regression analysis. In that analysis, both Canada and Mexico estimate the impact of the COOL measure not on the actual export price of livestock, but on the price basis, namely the difference or gap between the US price and their own export price (which is defined differently by Canada and Mexico).69 However, in respect of feeder pigs, Canada does not use regression analysis to estimate the price, but relies instead on a descriptive comparative analysis of prices based on invoices provided by a large Canadian firm trading both in the US domestic and the Canadian export market.70
3.7.
In respect of the estimation of quantity impacts, Canada's and Mexico's approaches differ. Canada estimates export quantities in the same way as price, namely through econometric estimation. Mexico employs an elasticity-based simulation using the estimated price impact and a derived elasticity figure to arrive at the impact on export quantities.71
3.8.
To calculate domestic price suppression losses, Canada and Mexico multiply the quantity of livestock in their respective domestic markets (subtracting the quantity exported in order to avoid double-counting) by the above counterfactual export price. Canada assumes full (one to one) transmission of export price effects to the domestic price. Mexico applies a transmission coefficient of 0,678 to account for factors that mitigate full transmission.72
3.9.
Canada's and Mexico's explanations of their methodologies, as well as details on their respective specifications and data used, are discussed in section 5 below.

3.2 THE RELEVANT COUNTERFACTUAL

3.10.
As noted above, Canada and Mexico compare actual revenues obtained in a given baseline year after the expiry of the RPT, and estimated revenues that would have been obtained in that year absent the COOL measure. Both Canada and Mexico base this comparison on the assumption that the United States would have come into compliance with the recommendations and rulings of the DSB by withdrawing the COOL measure.
3.11.
While the United States points out that there could be other options for compliance, it accepts, for the purposes of this arbitration, the counterfactual in which the COOL measure has been withdrawn.73 In its own alternative methodology, the United States applies this counterfactual to assess, within the baseline period of 2014, "the effect of removing any incentives or 'discounts' resulting from the amended COOL measure".74
3.12.
Thus, the parties are generally in agreement that the relevant counterfactual for the purposes of this arbitration is the situation that would exist in the baseline period in the absence of the COOL measure. The parties differ, however, in their approach to assessing the impact of the COOL measure.

3.3 THE UNITED STATES' THREE-PRONGED CHALLENGE AGAINST THE LEVELS OF SUSPENSION PROPOSED BY CANADA AND MEXICO

3.13.
The United States challenges the proposed levels of suspension in three separate ways, arguing that "[a]ny one of these ways is sufficient to meet the U.S. burden and each one on their own establishes that Canada and Mexico's requests are inconsistent with the DSU."75
3.14.
First, the United States directly challenges various aspects of the methodologies used by Canada and Mexico. The United States' main focus is on the use of econometric modelling, which it considers "inappropriate for the question at issue".76 More specifically, the United States contests the use of price basis and argues that Canada's and Mexico's econometric estimations are formulated incorrectly and suffer from variable omission. The United States also challenges Canada's feeder pig price estimation as well as Mexico's quantity simulation. Finally, the United States raises a number of specific issues relating to the data used to estimate COOL impacts under the econometric approach.
3.15.
Second, the United States also challenges the proposed level of suspension by submitting an alternative methodology. According to the United States, this alternative methodology "is appropriate for the question presented and accurately estimates the levels of nullification or impairment, as opposed to the econometric models proposed by the requesting parties."77 The alternative methodology proposed by the United States is a partial equilibrium model, more specifically an "equilibrium displacement model" (EDM). Applying its EDM, the United States calculates that Canada's export revenue losses amount to USD 43.22 million and Mexico's losses amount to USD 47.55 million.78
3.16.
Third, the United States challenges, as a threshold matter of legal interpretation and for methodological reasons, the inclusion of domestic price suppression losses in the level of nullification or impairment of benefits.79

3.4 CANADA'S AND MEXICO'S ARGUMENTS ON THE EDM

3.17.
Canada and Mexico contest the use of an alternative methodology as a means of setting out a prima facie case against their own methodologies.80 Canada and Mexico also raise a number of arguments against the methodology itself. Canada's and Mexico's main criticism in this respect is that the EDM proposed by the United States does not reflect the segregation and differential compliance costs which underlie the findings of WTO-inconsistency in this dispute.81 Furthermore, Canada and Mexico take issue with the elasticity values used by the United States.82 We describe how we address these arguments in section 4.3 below.

4 PRELIMINARY ISSUES

4.1 MANDATE OF THE ARBITRATOR

4.1.
The United States objects to the levels of suspension indicated by Canada and Mexico in their requests to the DSB on the grounds that these levels are not equivalent to the nullification or impairment caused.83 We begin by recalling our mandate as set out in Article 22.7 of the DSU, which states in relevant part:

The arbitrator acting pursuant to paragraph 6 shall not examine the nature of the concessions or other obligations to be suspended but shall determine whether the level of such suspension is equivalent to the level of nullification or impairment.84

4.2.
Thus, our task in these proceedings is to examine whether there is equivalence between the proposed level of suspension and the level of nullification or impairment.85 The nullification or impairment in question is, as the arbitrator in US – 1916 Act (Article 22.6 – EC) noted, that "sustained by the complaining party as a result of the failure of the responding party to bring its WTO-inconsistent measures into compliance".86
4.3.
"Equivalence", as the arbitrator in EC – Bananas (US) (Article 22.6 – EC) observed, "connotes a correspondence, identity or balance between two related levels, i.e. between the level of the concessions to be suspended, on the one hand, and the level of the nullification or impairment, on the other".87
4.4.
The levels of suspension that Canada and Mexico propose correspond to the levels of nullification or impairment that each has identified in their respective methodology papers. Our task is to assess Canada's and Mexico's determinations of their respective levels of nullification or impairment. If we cannot accept Canada's and Mexico's determinations, our mandate requires us to make our own determination.88 As the arbitrator in EC – Hormones (US) (Article 22.6 – EC) stated:

There is … a difference between our task here and the task given to a panel. In the event we decide that the US proposal is not WTO consistent, i.e. that the suggested amount is too high, we should not end our examination the way panels do, namely by requesting the DSB to recommend that the measure be brought into conformity with WTO obligations. Following the approach of the arbitrators in the Bananas case … we would be called upon to go further. In pursuit of the basic DSU objectives of prompt and positive settlement of disputes we would have to estimate the level of suspension we consider to be equivalent to the impairment suffered. This is the essential task and responsibility conferred on the arbitrators in order to settle the dispute. In our view, such approach is implicitly called for in Article 22.7.89

4.5.
We note that in making their own determination of the level of nullification or impairment, previous arbitrators developed their own appropriate methodologies90, which were either based on elements of the methodologies initially proposed by the parties91 or which followed an altogether different approach.92 We observe that any determination of nullification or impairment, because it is based on assumptions, is necessarily a "reasoned estimate" relying on "credible, factual, and verifiable information".93
4.6.
Our decision will determine the level of nullification or impairment with which the level of suspension shall be equivalent. We note that this level of suspension will represent the upper limit of any suspension of concessions or other obligations that Canada or Mexico may apply. While our decision, in this manner, allows the DSB to ensure "equivalence" in any authorization it grants in accordance with Article 22.4 of the DSU, subsequently it will be for the authorized Member to ensure that the suspension is applied in a manner that does not exceed the authorized level.94

4.2 BURDEN OF PROOF

4.7.
We agree with previous arbitrators on the applicable standard on burden of proof, which has been summarized by the arbitrator in EC – Hormones as follows:

WTO Members, as sovereign entities, can be presumed to act in conformity with their WTO obligations. A party claiming that a Member has acted inconsistently with WTO rules bears the burden of proving that inconsistency. The act at issue here is the US proposal to suspend concessions. The WTO rule in question is Article 22.4 prescribing that the level of suspension be equivalent to the level of nullification and impairment. The EC challenges the conformity of the US proposal with the said WTO rule. It is thus for the EC to prove that the US proposal is inconsistent with Article 22.4. Following well-established WTO jurisprudence, this means that it is for the EC to submit arguments and evidence sufficient to establish a prima facie case or presumption that the level of suspension proposed by the US is not equivalent to the level of nullification and impairment caused by the EC hormone ban. Once the EC has done so, however, it is for the US to submit arguments and evidence sufficient to rebut that presumption. Should all arguments and evidence remain in equipoise, the EC, as the party bearing the original burden of proof, would lose.95

4.8.
The same arbitrator also observed that "the same rules apply where the existence of a specific fact is alleged", noting that "[i]t is for the party alleging the fact to prove its existence."96
4.9.
Finally, as has been emphasized in previous arbitrations, all parties have a duty to produce evidence and to collaborate in presenting evidence to the arbitrator.97 It is this duty that requires a requesting party to submit a methodology paper "explaining how it arrived at its proposal and showing why its proposal is equivalent to the trade impairment it has suffered".98
4.10.
As seen above, one of the three ways in which the United States challenges the proposed level of suspension is by using a completely different alternative methodology, which it considers more appropriate and which results in a much lower level of nullification or impairment.99 We therefore see a need in this case to set out additional considerations on the legal standard of burden of proof, and in particular, to further explore the role of opposing methodologies submitted in Article 22.6 arbitration proceedings. In the United States' view, presenting a different calculation of the level of nullification or impairment is a prima facie demonstration that the levels proposed by Canada and Mexico are inconsistent with the DSU. While alternative methodologies have been proposed and discussed in previous arbitrations, this is the first time that an objecting party explicitly presents an alternative methodology on its own merits for purposes of satisfying its initial burden of proving that the level of nullification or impairment proposed is WTO-inconsistent.100
4.11.
The methodology papers submitted by Canada and Mexico respond to their duty, described above, to produce evidence and collaborate in presenting evidence to the arbitrator. Methodology papers are different from the actual request to suspend concessions or other obligations at a proposed level, which, as seen above, is the "act at issue" that is presumed to be in conformity with WTO obligations. However, the underlying methodologies are inextricably linked with the proposed level of suspension in that they substantiate and explain the grounds on which the act at issue is based. Because the proposed level of suspension rests on the underlying methodology, establishing that the proposed level of suspension is WTO-inconsistent necessarily involves showing that it does not follow from the underlying methodology, or that the methodology itself is flawed. This necessitates engagement by the objecting party with the methodology underlying the proposed level of suspension.
4.12.
It may be possible to present an alternative methodology as a way of engaging with, and contributing to disproving, a proposed methodology. However, merely putting forward, as was done here, a different methodology as "appropriate"101 or as one that "more accurately estimates"102 the level of nullification or impairment is not sufficientIn the absence of a demonstration that the proposing party's methodology is incorrect, the mere submission of an alternative methodology would not meet the objecting party's burden of proof. This is because the alternative methodology does not, in itself, assist the Arbitrator in determining whether the result from the first methodology is (or is not) equivalent to the level of nullification or impairment. In such a situation, it would follow from the rules on burden of proof that the objecting party has not proved that the act at issue is WTO-inconsistent.
4.13.
The onus is therefore on the United States to show that the proposed level of suspension is inconsistent with the DSU by engaging with the methodologies proposed by Canada and Mexico, and demonstrating that they do not lead to a result that is equivalent to the level of nullification or impairment.
4.14.
In sum, we are of the view that, in order to meet its prima facie burden, an objecting party under Article 22.6 of the DSU must engage with the methodology used to arrive at the proposed level of suspension and that it is not sufficient merely to assert that another methodology is more appropriate. We therefore find that, in merely proposing an alternative methodology, the United States has not validly established a prima facie case against the levels of suspension proposed by Canada and Mexico.

4.3 ORDER OF ANALYSIS

4.15.
In light of the above considerations on our mandate and the apportioning of the burden of proof, we proceed with our analysis in the following order.
4.16.
We will first assess the methodologies proposed by Canada and Mexico in examining whether the United States has successfully established that the proposed levels of suspension are in excess of the level of nullification or impairment. This assessment will focus on determining specific points of validity or error in the proposed methodologies based on the arguments and evidence submitted by the parties.
4.17.
In assessing the Canadian and Mexican methodologies, we will begin by considering which losses can be included in the nullification or impairment of benefits accruing to Canada and Mexico, and will then proceed to assessing the calculation of such losses. We observe that the United States, in its third line of argument described above, argues that losses from domestic price suppression cannot be included in the nullification or impairment considered under Article 22 of the DSU. As this involves a threshold question of a legal nature, we consider it appropriate, indeed necessary, to examine the permissible scope of relevant losses before turning to the actual calculations.
4.18.
We will then examine the actual calculations of the losses as presented by Canada and Mexico under their proposed methodologies. We will assess these calculations in light of the criticisms submitted by the United States in its first line of argument described above. We note that many of the arguments discussed in this respect turn on factual allegations that are contested between the parties, and we recall that each party bears the burden of substantiating its own factual allegations. While recognizing that our mandate under Article 22.6 of the DSU differs from that of panels, we will be guided by the principles of Article 11 of the DSU in objectively assessing the arguments made and evidence submitted by the parties.103
4.19.
We will examine all elements of Canada's and Mexico's methodologies in determining whether the proposed level of suspension is equivalent to the level of nullification of impairment. The reason is that, for the purposes of making our own determination, we will consider all elements of methodologies that are on the table, retaining those elements of the proposed methodologies that we conclude are acceptable. Likewise, we will consider the United States' EDM to assess its comparative merits and shortcomings, and to ascertain which elements if any of the EDM may assist us in deciding upon the approach to adopt for our reasoned estimate of the level of nullification or impairment.

5 ASSESSMENT OF THE PROPOSED LEVEL OF SUSPENSION

5.1 INCLUSION OF DOMESTIC PRICE SUPPRESSION LOSSES IN THE LEVEL OF NULLIFICATION AND IMPAIRMENT

5.1.
In this section, we address the United States' challenge against the inclusion of domestic price suppression losses in Canada's and Mexico's determination of the level of nullification or impairment of benefits. We recall that Canada claims domestic price suppression losses in the amount of CAD 1,032.1 million. Mexico claims domestic price suppression losses in the amount of USD 198.6 million.104

5.1.1 Arguments of the parties

5.2.
The United States objects to the inclusion of such losses on the grounds that there is "no basis under the DSU for considering domestic price suppression as a part of the level of nullification or impairment of benefits under the TBT Agreement or the GATT 1994".105 The United States asserts that the level of nullification or impairment flows from the benefits under the covered agreements106, and in the present case the trade benefit "relates to international trade in livestock, not to domestic markets."107 The United States argues that past Article 22.6 arbitrations "involving the Multilateral Agreements on Trade in Goods have focused on the 'trade effect' of the WTO-inconsistent measure".108 According to the United States, Canada's and Mexico's "claims with respect to internal transactions within their domestic economies … are not lost exports to the United States, and thus are not properly included in a measurement of either Canada or Mexico's nullification or impairment of trade benefits under the covered agreements".109 The United States also argues that, if domestic price suppression losses are included in the nullification or impairment, the level of suspension would similarly have to account for broader economic effects of the suspension within the United States in order to maintain equivalence under Article 22.4 of the DSU.110 In the United States' view, "the requesting parties' approach would include domestic price effects on only one side of the equation (the side that benefits them), and would omit it and other economic effects from the other side of the equation" in contravention of the equivalence requirement.111
5.3.
According to Canada, Article 3.3 of the DSU "sets out a very broad ground rule for WTO dispute settlement" that includes benefits accruing "directly or indirectly" under the covered agreements.112 Canada defines the benefit in question as "national treatment for Canadian live cattle and hogs in the United States", a benefit that was adversely affected resulting, given the "highly integrated and co-dependent nature of the two markets … in more Canadian livestock in Canada, which suppressed the prices of these animals in the Canadian market, resulting in specific and quantifiable losses."113 Canada maintains that these are "direct losses from the denial of a direct benefit".114 Canada argues in the alternative that these domestic price suppression losses are "at the very least losses that result from the impairment of an indirect benefit of national treatment, which is a benefit covered under the DSU".115 Canada submits that there is nothing in the DSU that limits the level of nullification or impairment to "export losses", and cites prior Article 22.6 arbitrations in support of "a broad interpretation of 'trade effects'."116 Canada thus contends that "trade effects" need not be limited to export losses, but can include domestic impacts where causation can be demonstrated.117
5.4.
Mexico states that the benefit being nullified or impaired is the "right of not having to face a measure like the COOL measure."118 According to Mexico, it is "[b]y virtue of the nullification or impairment of this benefit by the COOL measure, [that] Mexican domestic prices have been suppressed."119 Mexico argues that the covered agreements refer to direct or indirect benefits, and that Mexico's benefits under the covered agreements "should have prevented this [domestic] price suppression from occurring".120 Mexico also relies on previous arbitrations in which it contends effects on domestic markets were not excluded.121 Mexico thus submits that the losses to be calculated in estimating nullification or impairment are those that can be shown to be caused by the WTO-inconsistent measure.122

5.1.2 Analysis by the Arbitrator

5.5.
The question raised in this arbitration is whether (and, if so, how) "price suppression losses" incurred by Canadian and Mexican livestock producers in their domesticmarkets can be included in the level of nullification or impairment under Article 22 of the DSU.
5.6.
Although prior arbitrators have considered losses other than those based strictly on actual trade flows, this specific question has not previously been addressed in arbitration proceedings. For instance, the arbitrator in EC – Bananas III (Ecuador) (Article 22.6 – EC) rejected Ecuador's argument that "the total economic impact of the EC banana regime should be taken into account by the Arbitrators by applying a multiplier when calculating the level of nullification and impairment suffered by Ecuador", on the grounds that Ecuador had not included this in its initial request to the DSB.123 The arbitrator in US – Gambling (Article 22.6 – US) similarly decided not to apply "a multiplier reflecting the aggregate change in output" and indirect, cross-sectoral effects of the measure on the domestic economy, but this decision was not based on a legal interpretation of the scope of "benefits" accruing under the covered agreements.124 While the arbitrators in US – 1916 Act (EC) (Article 22.6 – US) and US – Section 110(5) Copyright Act (Article 25) did not strictly limit their analysis to "trade effects", their "reliance on the broader concept of economic impact was dictated by the nature of the measures at issue", which did not directly restrict trade.125 Moreover, the effects they considered were not focused on economic gains or losses within the domestic market of the requesting parties.126 Finally, although the arbitrator in US – Offset Act (Byrd Amendment) (Article 22.6) observed that "the term 'trade effect' is found neither in Article XXIII of the GATT 1994, nor in Article 22 of the DSU", it recognized that "the 'trade effect' approach … seems to be generally accepted by Members as a correct application of Article 22 of the DSU".127
5.7.
As discussed above, our mandate under Article 22.7 of the DSU is to "determine whether the level of … suspension is equivalent to the level of nullification or impairment". As established by Article 3.1 of the DSU, the concept of "nullification or impairment" is taken from Article XXIII:1 of the GATT 1994, which provides for the nullification or impairment of "any benefit accruing […] directly or indirectly under this Agreement".128
5.8.
Neither "nullification or impairment" nor the "benefit" accruing under the covered agreements is explicitly defined in the GATT 1994 or in the DSU. The Appellate Body commented on the scope of these concepts in the context of Article 3.8 of the DSU, which provides:

In cases where there is an infringement of the obligations assumed under a covered agreement, the action is considered prima facie to constitute a case of nullification or impairment. This means that there is normally a presumption that a breach of the rules has an adverse impact on other Members parties to that covered agreement, and in such cases, it shall be up to the Member against whom the complaint has been brought to rebut the charge.129

5.9.
The Appellate Body observed that "Article 3.8 equates the concept of 'nullification or impairment' with 'adverse impact on other Members', although the DSU does not define 'adverse impact'".130 The Appellate Body further considered that "[t]rade losses represent an obvious example of adverse impact under Article 3.8."131 At the same time, the Appellate Body did not purport to provide a comprehensive explanation of the types of adverse impact that can be presumed in the case of WTO-inconsistent measures; nor was the Appellate Body concerned with quantifying nullification or impairment as is our mandate to fulfil as Arbitrator under Article 22 of the DSU.
5.10.
In the context of non-violation complaints under Article XXIII:1(b) of the GATT 1994, "the claimed benefit has been considered to be that of "legitimate expectations of improved market-access opportunities arising out of relevant tariff concessions."132 The "nullification or impairment" of such benefits has been equated with "'upsetting the competitive relationship' established between domestic and imported products as a result of tariff concessions".133 In the compliance phase of these disputes, the panel applied a similar understanding of the "nullification or impairment of benefits" with respect to Canada's and Mexico's non-violation claims under Article XXIII:1(b) of the GATT 1994 and Article 26.1 of the DSU.134 Although the panel was addressing a distinct issue in that context135, the applicable principles for non-violation nullification or impairment suggest that market access is the primary, though possibly not exclusive, benefit that is nullified or impaired. Such market access may be impaired not only by violations of tariff concessions but also by violations of rules and disciplines on non-tariff measures.136
5.11.
Unlike non-violation claims, this arbitration concerns the nullification or impairment of benefits that flow from specific provisions violated by the COOL measure.137 The link between the WTO-inconsistency and the benefits that are nullified or impaired is evident in the text of Article XXIII:1(a) of the GATT 1994, which stipulates that nullification or impairment occurs "as a result of the failure of another contracting party to carry out its obligations".138 Thus, the relevant "benefits" being nullified or impaired are those accruing to Canada and Mexico under the provisions breached by the COOL measure, namely the national treatment obligations of Article 2.1 of the TBT Agreement and of Article III:4 of the GATT 1994.
5.12.
It is well-established that the national treatment obligation that has been infringed "requires effective equality of opportunities for imported products to compete with like domestic products".139 At least one benefit flowing from national treatment, therefore, is the ability to compete in a foreign market, which in this case means market access for livestock imported into the United States.140 Where such market access is the benefit that is being nullified or impaired, the quantification of that nullification or impairment will naturally focus on trade flows (or a proxy thereof) as the measure of such access.141 We note that Canada's and Mexico's calculations of lost export revenues are aimed at doing just that, namely quantifying the amount of lost market access.
5.13.
By additionally claiming losses from domestic price suppression, Canada and Mexico go beyond the concept of market access and "trade effects" as the measure of market access. The question, therefore, is whether, in the context of determining nullification or impairment under Article 22 of the DSU, the benefits flowing from national treatment go beyond the benefit of market access, and particularly whether they extend to price effects in the domestic market of a requesting party.
5.14.
Canada and Mexico submit that the benefits do go beyond market access, essentially by understanding "nullification or impairment of benefits" to refer to any adverse effects resulting from the violation of the national treatment obligations at issue. According to this logic, the determinative criterion for including or excluding losses would be the causal link between the violation and the claimed effect.142 We disagree with this view for the three reasons set out below.
5.15.
First, with regard to the ordinary meaning of the relevant terms, the breadth of the term "benefit" as used in the covered agreements does not mean that it is unlimited. We recall that the compliance panel in these disputes set forth the following considerations for the scope of the term "benefits" and the significance of their accrual "directly or indirectly":

Article XXIII:1(b) of the GATT 1994 refers to "any benefit accruing to [a Member] directly or indirectly under [the GATT 1994]." Similarly, Article 26.1 of the DSU refers to "any benefit accruing to [a Member] directly or indirectly under the relevant covered agreement". Dictionary definitions of "benefit" include "an advantage, a good" and "pecuniary profit".143 Additionally, dictionary definitions of "accrue" include "of a benefit or sum of money [, to] be received in regular or increasing amounts" and "arise or spring as a natural growth or result".144 In principle, these definitions do not preclude that a benefit may "accrue" without being actually utilized.

By protecting benefits that accrue "directly or indirectly", both Article XXIII:1(b) of the GATT 1994 and Article 26.1 of the DSU suggest a possibly broad scope for the term "benefit". Further, both Articles refer to "any" benefit. Given the dictionary definition of the word "any", these provisions might apply "no matter which, or what"145 particular benefit is at issue. This would not support narrowing the term "benefit" to a specific manner of enjoyment or entitlement.146

5.16.
The foregoing examination of the ordinary meaning of the relevant terms – albeit under separate provisions regarding non-violation claims – is indicative of the potential breadth of the benefits accruing under the covered agreements. However, this in itself does not answer the specific question of whether the claimed domestic losses are within the scope of benefits that are nullified or impaired by a WTO-inconsistency. Even under this broad definition, a "benefit" is an "advantage" that is received (or legitimately expected), and it is this "advantage" that is being nullified or impaired. The benefit that is nullified or impaired, thus, is conceptually distinct from the right from which it flows.147 Canada and Mexico, in describing the benefit as "the national treatment for Canadian live cattle and hogs in the United States"148 and "the right of not having to face a measure like the COOL measure"149, effectively equate right with benefit. As we see it, the right in question is for imported products not to receive less favourable treatment than domestic products; the extent to which the advantage flowing from the right has been diminished is a separate question from what that right is. Thus, the right to national treatment under the covered agreements does not itself establish or prejudge the scope of benefits accruing therefrom.
5.17.
Moreover, we do not consider the phrase "directly or indirectly" to be a clear basis for distinguishing benefits accruing "directly" and those accruing "indirectly" so as to differentiate which losses are part of the nullification or impairment. Although the integral phrase "directly or indirectly" weighs against a narrow reading of "benefits", this does not necessarily extend the scope of nullification or impairment to other losses such as those caused, as is claimed here, by domestic price suppression. Indeed, both Canada and Mexico submit that domestic price suppression losses constitute the nullification or impairment of a benefit directly accruing to them, and in the alternative claim that such losses correspond to benefits accruing indirectly.150
5.18.
Second, in terms of relevant context, we see a number of contextual provisions within the DSU as well as the SCM Agreement that weigh against reading "nullification or impairment of benefits" in the manner suggested by Canada and Mexico. We consider this context in interpreting the provisions of the WTO covered agreements in a coherent manner, giving meaning to all provisions harmoniously.151 Articles 21.8 and 22.3(d)(ii) of the DSU, which are immediate context to Article 22.7, suggest that the consideration of domestic economic effects is distinct from measuring the nullification or impairment of benefits. Article 21.8 of the DSU applies to cases brought by developing country Members, and directs the DSB to "take into account" the "impact on the economy of developing country Members concerned". This provision (which has not been raised in these proceedings as a basis for including domestic price suppression losses) does not address the level of nullification or impairment that it is our mandate to assess under Article 22 of the DSU. In particular, the text of this provision suggests that it relates to a requirement imposed on the DSB to take into account specific factors "in considering what appropriate action might be taken". This does not concern arbitration under Article 22.6, but rather the DSB's discharge of its functions in Article 2.1 of the DSU regarding "the surveillance of implementation of DSB rulings and recommendations" that is the subject of Article 21 of the DSU.
5.19.
Article 22.3(d)(ii) of the DSU addresses "principles and procedures" that complaining parties are required to apply among others "[i]n considering what concessions or other obligations to suspend", and provides that a "party shall take into account … the broader economic elements related to the nullification or impairment and the broader economic consequences of the suspension of concessions or other obligations". Crucially, this provision is relevant when assessing a request to cross-retaliate (i.e. across different sectors and agreements than those in which violations were found) and thus concerns the specific targets of the suspended concessions. Importantly for our analysis, it does not concern the level of that suspension based on the nullification or impairment of benefits.152 Thus, while the DSU provides for consideration of domestic economic effects in specific contexts, it makes no indication of similar considerations being relevant to the level of nullification or impairment that it is our mandate to assess.
5.20.
Furthermore, we note that the SCM Agreement makes it clear that "nullification or impairment" is a concept that is distinct from other adverse effects and, in particular, from domestic injury. Article 5 of the SCM Agreement sets out three distinct categories of "adverse effects", namely (a) injury to the domestic industry, (b) nullification or impairment, and (c) serious prejudice. Nullification or impairment is explicitly linked to the GATT 1994, with footnote 12 to Article 5 of the SCM agreement stipulating that "[t]he term 'nullification or impairment' is used in this Agreement in the same sense as it is used in the relevant provisions of GATT 1994, and the existence of such nullification or impairment shall be established in accordance with the practice of application of these provisions." Article 5 also makes clear that injury to domestic industry specifically encompasses the effect of "depress[ing] prices to a significant degree" or preventing price increases.153 We consider it meaningful for our findings that the SCM Agreement explicitly distinguishes such domestic price suppression effects from nullification or impairment in the sense of the GATT 1994.154
5.21.
Third, in addition to the contextual arguments above, we consider the preamble to the WTO Agreement, which the parties discussed at the substantive meeting. To the extent that the preamble sets out the "objectives" of the treaty, an initial point is that the term "objectives" is not to be conflated with the term "benefits". This is readily apparent from Article XXIII:1 of the GATT 1994, which refers separately to situations in which "any benefit … is being nullified or impaired" and those in which "the attainment of any objective of the Agreement is being impeded". We note that Article 22 of the DSU does not contain any reference to the objectives of the covered agreements being impeded, but only to nullification or impairment; by contrast, Article 26 of the DSU concerning non-violation and situation complaints is addressed to nullification or impairment or the attainment of any objective being impeded. Thus, the fact that domestic price suppression caused by a WTO-inconsistency may impede certain objectives of the Agreement does not mean that such price suppression is the nullification or impairment of a benefit under Article 22 of the DSU.
5.22.
The preamble to the WTO Agreement makes clear in its first recital that trade relations are linked to domestic economic gains such as "raising standards of living, ensuring full employment and a large steadily growing volume of real income and effective demand, and expanding the production of and trade in goods and services".155 In addition, the third recital of the preamble expresses the desire "of contributing to these objectives by entering into reciprocal and mutually advantageous arrangements directed to the substantial reduction of tariffs and other barriers to trade and to the elimination of discriminatory treatment in international trade relations".156 Similarly, the fourth recital of the preamble expresses the fundamental resolution "to develop an integrated, more viable and durable multilateral trading system" on the basis of past trade liberalization and the results of prior trade negotiations. Thus, while the economic gains ultimately derived from trade are not limited to trade flows themselves, the WTO Agreement frames such broader economic gains as an end for which trade and market access are an essential means.
5.23.
It is the interests of trade and market access that underlie Members' concessions and the legal remedies designed to safeguard those concessions. Hence, a "major goal for [Articles XXII and XXIII of the GATT 1994] was to provide a means for ensuring continued reciprocity and balance of concessions in the face of possibly changing circumstances."157 Insofar as the rights and obligations under the covered agreements are grounded in a balance of trade concessions, the nullification or impairment of benefits is appropriately focused on the trade losses occasioned by a disruption of those concessions. Simply put, trade is a means to broader economic gains, and this trade is protected as a benefit accruing under the covered agreements.
5.24.
In light of the foregoing, the fact that adverse effects may exist beyond trade losses does not necessarily imply their inclusion in the level of "nullification or impairment of benefits" under Article 22 of the DSU. Indeed, it is readily conceivable that trade losses would result in corresponding domestic impacts – just as the trade disciplines of the WTO agreements are expected to foster domestic economic gains, such gains may be diminished or lost when there is a violation of those disciplines. This does not mean that all such losses may be rebalanced through suspension of concessions under Article 22 of the DSU. In the case of domestic price suppression, identifying the "net loss" suffered would raise an additional question of whether, and how, to account for positive effects for downstream consumers of the price-suppressed product in the domestic market.
5.25.
At the same time, we are not persuaded that including domestic price suppression would require us to account for similar economic impacts in the domestic market of the objecting Member in order to ensure "an apples-to-apples determination of equivalency".158 The equivalence requirement set out in Article 22.4 of the DSU applies to the equivalence of two levels, irrespective of how the concepts of "nullification or impairment" and "benefit" are interpreted. It is not for the Arbitrator to ascertain whether equivalence is maintained in the application of countermeasures. This would require the measurement of future losses as well as an examination of the nature of the concessions suspended, which Article 22.7 expressly prohibits us from considering.159 Should the effect of the suspension of concessions exceed the level of nullification or impairment, whether due to the manner of application or the nature of concessions suspended, the Member concerned could have recourse to DSU procedures challenging the consistency of the level of the suspension with Article 22.4 of the DSU.160
5.26.
Finally, all parties have referred to the proposal being considered by Members in the DSB Special Session as part of the "negotiations on improvements and clarifications" to the DSU.161 This proposal concerns an amendment of Article 22.4 of the DSU to take into account the economic impact of the inconsistent measure on the economy of a developing country complainant.162 Canada and Mexico note the differences between the original proposal under negotiation (regarding broader economic effects on developing countries) and the nature of their requests, which pertain to direct losses to a particular domestic industry.163 We agree with Canada and Mexico that the proposal being discussed in the DSU negotiations provides little interpretive guidance for the question presented in this arbitration, including due to the substantive differences between the proposal itself and the particular losses claimed by Canada and Mexico. More generally, we are not persuaded by the United States' contention that negotiation of a given item necessarily proves that it does not exist under current DSU rules.164 Proposals to clarify and improve existing DSU rules are without prejudice to Members' differing views165 on the legal interpretation of the rules as they currently stand.
5.27.
In conclusion, we consider that the relevant benefit in this case is the market access that has been nullified or impaired as a result of the COOL measure. Therefore, we do not include the domestic price suppression losses claimed by Canada and Mexico in the level of nullification or impairment of benefits. Consequently, we focus the remainder of our analysis on the claimed level of export revenue losses caused by the COOL measure.

5.2 CALCULATION OF LOST EXPORT REVENUES

5.28.
In this section, we assess the proposed level of suspension by reference to the level of nullification or impairment caused by the COOL measure, as calculated by Canada and Mexico in respect of export revenue losses. As noted above in section 3.1, Canada submits CAD 2,045 million and Mexico USD 514.8 million in export revenue losses. To calculate the export revenue losses caused by the COOL measure, Canada and Mexico separately estimate impacts on export prices and quantities, which we address accordingly in separate sections below.
5.29.
We observe that Canada and Mexico principally rely on econometric analysis, specifically linear regression analysis, in their respective methodologies for calculating the impact of the COOL measure. As noted above, in addition Canada relies on a descriptive analysis to estimate the impact of COOL on prices for feeder pigs, and Mexico relies on an elasticity simulation to estimate the impact of COOL on quantities of feeder cattle exported to the United States.166
5.30.
As regards econometric analysis, we recall that the original and compliance panels in these disputes examined economic and econometric evidence submitted in connection with the legal claims raised in respect of the original and amended COOL measure's detrimental impact on imported livestock. As stated by both panels, it was not necessary to verify actual trade effects to dispose of the national treatment claims before them, and the review of such evidence was pursuant to the function of panels to make an objective factual assessment under Article 11 of the DSU. Further, the original panel emphasized that this assessment did not concern "any level of nullification or impairment, let alone whether there is any equivalence with any suggested level of suspension of concessions or other obligations", as these were matters "to be decided by an eventual arbitration under Article 22.6 of the DSU and on the basis of evidence submitted in the context of such arbitration".167
5.31.
Given the prominence of econometrics in these proceedings, we briefly set out a background explanation of the main features of this methodology. Essentially, in a linear regression analysis, a "dependent" variable (that is, the variable of interest) is modelled as a linear function of a number of "explanatory" variables. These explanatory variables ideally represent the full set of factors that have an impact on the dependent variable, and therefore contribute to "explaining" the behaviour of the dependent variable. In general, the explanatory variables are assumed to be independent with respect to the dependent variable. In other words, the dependent variable is assumed to have no impact (direct or indirect) on the explanatory variables that, in turn, have an impact on the dependent variable.
5.32.
For each explanatory variable included in the econometric model, a specific parameter is attached to it. This parameter represents the impact that the associated explanatory variable might have on the dependent variable. Thus, when the econometric model is well specified with all relevant explanatory variables, each parameter isolates the impact of the associated explanatory variable on the dependent variable.168 In their methodologies, Canada and Mexico use parameters associated with the COOL measure to calculate what the export price (and quantity, for Canada) of livestock would have been without the combined effect of the original and amended COOL measure.
5.33.
While we note the United States' broad contention that econometric modelling is unsuitable for accurately estimating the impact of the COOL measure, principally due to the alleged impossibility of accurately accounting for all relevant variables, we also note that arguments regarding the fundamental flaws or unsuitability of econometric modelling to estimate trade effects cannot be assessed in the abstract.169 We therefore assess Canada and Mexico's proposed levels of suspension by examining the specific application of their methodologies in determining the level of nullification or impairment caused by the COOL measure.

5.2.1 Price impact estimation

5.2.1.1 Use of price basis to estimate the COOL impact on export prices

5.34.
As stated in section 3.1, for their computation of export revenue loss, Canada and Mexico rely on an econometric estimation of the impact of the COOL measure on "price basis".170 The price basis is the differential between the price of Canadian/Mexican exported livestock and the price of US-origin livestock. Thus, Canada and Mexico do not directly estimate the impact of the COOL measure on the absolute level of export prices; rather, they equate the absolute price impact with the degree to which the COOL measure has widened the price basis (i.e. increased the difference between the export price and the US price) to the detriment of foreign-origin livestock.
5.35.
Before addressing the parties' arguments, we note that Canada and Mexico define the price basis differently. Canada defines the price basis as the difference between the export price of Canadian livestock in Canada and the price of United States' livestock in the United States171, while Mexico defines the price basis as the difference between the price of exported Mexican livestock in the United States and the price of United States' livestock in the United States.172 In other words, Canada compares the price of livestock in two different countries (i.e. Canada and the United States), while Mexico compares the price of livestock in the same country (i.e. the United States).
5.36.
In this Decision, reference to the "export price" means the price of Canadian and Mexican livestock as defined in their respective methodologies, unless specified otherwise. Additionally, reference to the "US price" means the price of comparable livestock of US-origin within the United States.

5.2.1.1.1 Arguments of the parties

5.37.
The United States challenges the Canada’s and Mexico’s use of the price basis, rather than the actual export price, to estimate lost export revenue.173 The United States argues that if Canada's and Mexico's "export equations had all the proper exogenous variables then [they] could have used those same exogenous variables to explain the [effect on Canadian and Mexican] prices directly rather than just through a price basis analysis."174 The United States notes, however, that because livestock prices have increased during the relevant period, applying the same exogenous variables that Mexico and Canada used in their analyses would show that the COOL measure actually caused higher prices.175 According to the United States, this demonstrates the "flaws" (in respect of Mexico), or "limited explanatory value" (in respect of Canada), of a price basis regression.176
5.38.
According to the United States, Canada's and Mexico's model specifications of the price basis also prevent them from distinguishing between the impacts of the COOL measure on the Canadian/Mexican livestock export price and on the United States' livestock price.177 The United States asserts that the price basis "includes more than the change in Canadian or Mexican price – it naturally also relates to fluctuations up and down in the U.S. price and the rate of these changes. This difference between two prices cannot be automatically equated with the price level for exports of Canadian and Mexican livestock."178 Because the widened price basis may be due to increased US prices, and not just declining import prices, "[u]sing the price basis for determining the actual trade impact of COOL will overstate the price effect."179
5.39.
The United States supports its assertion that the COOL measures increased the price of United States' livestock through reliance on economic logic, academic research, and data showing that "U.S. prices for U.S. origin livestock have consistently increased following the implementation of the COOL measures".180 The United States also refers to its own application of Canada's "price model specification and weekly data to review the U.S. price levels … [to show] that the impact of the original and amended COOL measures on the U.S. price is in fact positive."181
5.40.
In order to show that a change in the price basis is not equivalent to a change in actual price, the United States notes the definitional and mathematical distinction between the two concepts. Based on this distinction, the United States asserts that, "[i]n principle, any change in the U.S. price will result in a change in the price basis unless it is exactly offset by a change in the Canadian or Mexican export price."182 The United States provides data showing volatility in the Canadian-United States' price basis over time and suggests that "if Canada's argument that the basis (i.e., difference between the U.S. and Canadian) prices remained steady over time, until the amended COOL measure expanded the basis, we would expect the basis described by the line to be flat during this period, rather than wildly fluctuating."183 According to the United States, this makes it "clear [that] other causal factors have affected the basis, and they need to be accounted for in econometric modeling".184 The United States also argues that "sample econometric analysis conducted by the United States based on the equations and data provided by the requesting parties supports the understanding that price level change and price basis are not equivalent."185
5.41.
Canada argues that the use of a price basis specification "allows one to capture parsimoniously the impacts of a host of variables that may affect livestock prices in both countries in a similar way."186 Canada thus suggests that the use of a price basis obviates the need to account for any and every variable that might impact the price of livestock in the United States' and Canadian markets.187 Canada submits that the "positive theoretical impact of the amended COOL measure on U.S. price through reduced import competition will be small because the share of imports is so small", and that any "small positive impact of the domestic impacts of COOL on the U.S. price is countered by small negative effects on U.S. prices", as reported in a study of US domestic effects.188
5.42.
Canada also responds to evidence submitted by the United States to support the contention that a change in price basis is not equivalent to a change in export price levels. Canada notes that the United States' evidence of volatility in the price basis over time is "irrelevant" because "Canada has never taken the position that there are no fluctuations in the price basis … [and] Canada never argued that the basis remained steady over time before or after COOL."189 Canada contends that the United States' evidence provides "no guidance about the causation related to the amended COOL measure … [whereas] Canada's price basis regressions account for most of this variation and specifically isolate the causal effect of COOL on the price basis."190 Canada criticizes the United States' use of Canada's econometrics for "cherry-picking" a single animal category and for "mis-specifying" the equations.191 Canada also criticises the United States' "sample econometric analysis" for relying on parameters acknowledged to be flawed, and confusing the units of measurement.192
5.43.
Mexico submits that the "objective of Mexico's regression model is to explain how the differential treatment of cattle in the United States, according to their origin, affected the price paid for Mexico feeder cattle."193 Mexico asserts that the price basis measures "only the difference in value to the US feeding operations for feeder cattle of Mexican and US origins."194 The use of price basis rather than price explains the differential impact, especially because, "[w]ith prices measured in the same locations, the number of variables that affect the basis is limited."195 Mexico adds that, while the methodology could be applied to the actual "price paid", such a model "would be plagued with problems that the [price] basis regression does not have."196
5.44.
Regarding any potential COOL-related price increase in the United States, Mexico notes that such an "increase in … price … would be small in practice because the market share of imported cattle is small relative … to the total size of the U.S. domestic cattle and beef industry", and change in import volume caused by the COOL measure is an even smaller share of the US market.197 Additionally, according to Mexico, the arbitrage mechanism in the integrated Northern American livestock market ensures that the "difference between the two prices reflects exactly the costs associated with the COOL measure that [are] passed on to Mexican feeder cattle."198 Additionally, regarding the United States' evidence of non-equivalence between change in price basis and change in price, Mexico suggests that the regression model submitted by the United States for this purpose does not address Mexico's model199, and is "mis-specified" for failing to include US cattle prices as an explanatory variable and for incorrectly applying a first difference to the COOL variables.200 Mexico also argues that the short-run volatility in the price basis, as identified by the United States, is normal and is accounted for through Mexico's long run econometric regression.201

5.2.1.1.2 Analysis by the Arbitrator

5.45.
We begin by recalling the methodological background against which Canada and Mexico apply their estimates of the COOL impact on price basis. Canada and Mexico quantify the level of nullification or impairment as an expression of lost export revenue, which is defined as the export price (P) multiplied by the associated export volume (Q). It follows that a change (Δ) in export revenue is defined as the change in the product of export price and export volume (∆(PQ)). For the purpose of this arbitration, Canada and Mexico propose to compare two terms: (1) export revenue observed in the baseline period with the COOL measure in place and (2) export revenue that would have been obtained in the absence of the COOL measure. Thus, the difference between the revenue "with" and "without" the COOL measure represents the export revenue loss caused by the COOL measure. In this context, Canada and Mexico demonstrate that the expression of export revenue loss can be decomposed into three components, where P and Q represent export prices and quantities in absolute values in the baseline period, and ∆P and ∆Q represent the counterfactual change in export prices and quantities without the COOL measure202:

∆(PQ) = ∆PQ +∆QP - ∆P∆Q

5.46.
Canada and Mexico note that while data on the export price (P) and export quantities (Q) in absolute levels in the baseline period are readily available, there are no directly available data for the two differential terms, ∆P and ∆Q, measuring respectively the change in export price and export quantity between the baseline period with the COOL measure and a counterfactual situation without the COOL measure. Both Canada and Mexico propose to estimate separately the counterfactual change in export prices (∆P) and export volumes (∆Q) in the absence of the COOL measure.
5.47.
While the change in export price of livestock in absolute terms is one of the key components of the expression of export revenue losses, Canada and Mexico econometrically estimate the impact of the COOL measure on the price basis (rather than on the absolute price level). Both Canada and Mexico interpret the estimated coefficients of the COOL measure in the price basis specification as the impact of the COOL measure on the export price. In other words, the methodologies of Canada and Mexico rest on the assumption that the counterfactual impact of the COOL measure on the price basis is the same as the counterfactual impact on the Canadian/Mexican export price.
5.48.
As submitted by the United States, there is a basic definitional and mathematical difference between absolute export price levels and price basis differentials.203 For example, a change in price basis can be represented as follows:

Δ(US Price – Export Price) = Δ(US Price) – Δ(Export Price)

In support of its contention that the change in price basis should not be equated with a change in export price, the United States adduces the following illustration:

Δ(US Price) – Δ(Export Price) ≠ Δ(Export Price)

5.49.
This illustration reflects two key considerations about the use of price basis to estimate the impact of the COOL measure on export prices and, by extension, export revenues. One consideration concerns the irrelevance of variables that have the same impact on US and export prices. A second consideration concerns the conditions under which price basis analysis would yield an accurate measure of an explanatory variable's impact on the export price.
5.50.
Regarding the first consideration, it is clear that any variable that has an equivalent effect on the US price and the export price of livestock will have no effect on the price basis. In other words, if a given variable increases or decreases the export and US price by the same amount, the differential between the prices will remain the same. Indeed, this is a fundamental premise of Canada's and Mexico's defence of the use of price basis for the econometric estimation of the COOL impact on export price levels. Because price basis only changes when a variable differentially affects the export and US price, Canada and Mexico argue that the price basis equations only need to account for a limited number of variables that have such a differential impact.
5.51.
While it is mathematically apparent that variables with an equal impact on export and US prices do not affect the price basis, the implications of this fact in the present case relate to the specific context of the North American livestock market. As noted in the original proceedings of these disputes, the market for livestock (and meat) of Canada, Mexico, and the United States is highly integrated, with different stages of livestock and meat production often being performed in more than one country.204 Moreover, the vast majority of Canadian and Mexican livestock exports is destined for the United States, although imports from Canada and Mexico account for only a small percentage of total livestock slaughter in the United States.205
5.52.
A consequence of this integrated market is that livestock producers will sell their livestock wherever they are able to realize the highest return. In theory, such arbitrage conditions operate to establish a "law of one price", according to which a product is sold for the same price in all locations – when prices differ within the market, arbitrage operates to equalize the price difference between locations.206
5.53.
This is not to say that the markets of Canada, Mexico, and the United States are perfectly integrated. Indeed, the fact that there is a price differential between products of different origin indicates that there are certain factors leading to a departure from the theoretical "law of one price". The statistical volatility of the price basis observed by the United States does not itself contradict the notion that North American livestock markets are highly integrated. Rather, fluctuation in the price basis over time is consistent with the premise that there are certain factors differentially impacting livestock prices, and that market frictions may impede instantaneous adjustment to economic changes.207 At the same time, there is evidence that North American prices generally move together along the same trends, notwithstanding the existence of a differential between prices of livestock from different origins.208 For instance, the correlation between the US price and Canadian export price of livestock is extremely high and ranges between 0.95 and 0.98 for feeder/fed cattle, and is 0.98 for hogs.209 Similarly, the correlation between the US price and Mexican export price of feeder cattle is characterized by a high correlation of 0.98.210
5.54.
It is in this light that examination of changes in the price basis becomes relevant to assessing the impact of factors such as the COOL measure. In particular, price basis represents the price gap between livestock in the United States and comparable animals from Canada or Mexico that is due to trade costs, including transport costs and technical barriers to trade. For this reason, price basis analysis is a standard way of measuring trade costs caused by non-tariff measures.211 In addition, the key methodological advantage of focusing on price basis is to restrict the set of relevant variables to those that, like the COOL measure, have a differential impact on livestock prices.
5.55.
We turn to the second consideration regarding use of price basis to estimate the impact of the COOL measure on export prices, namely the conditions under which price basis analysis would yield an accurate measure of the change in export price. To accurately measure the negative COOLimpact on exports, we agree with the United States that the parameter estimates of the COOL measure should not capture any increase in US prices caused by the COOL measure. Any such increase would attribute a widening of the price basis to the COOL measure in calculating export revenue losses, even though this would not actually correspond to (and in fact would overstate) the negative impact on export prices. In this regard, we agree with the United States' assertion that, "[i]n principle, any change in the U.S. price will result in a change in the price basis unless it is exactly offset by a change in the Canadian or Mexican export price."212
5.56.
In this case, however, as we explain below, we do not find convincing evidence that the COOL measure led to increased US livestock prices. Therefore, we do not accept the United States' contention that price basis regression overestimates the reduction in export prices caused by the COOL measure. We note the United States' explanation that the COOL measure increased the price of United States' livestock based on the economic logic that "[t]he increased costs associated with the original and amended COOL measure result in decreased U.S. demand, as well as decreased Canadian and Mexican exports of livestock to the United States. This in turn results in an increase in the U.S. price of livestock."213 We find it useful to examine these contentions in the context of the findings adopted in previous stages of these disputes relating to the discriminatory impact of the COOL measure.
5.57.
With regard to the COOL impact on US demand, we recall the findings in the original and compliance stages regarding how the costs of the COOL measure are borne in the US market. The panels in both the original and compliance stages of these disputes found that the costs of the COOL measure could not be fully passed on to consumers, largely based on the USDA's own assessments that there was little evidence of consumer willingness to bear price increases commensurate with the added costs of mandatory labelling.214 The additional costs imposed by the COOL measure were thus largely passed up the supply chain to producers, for whom the least costly business scenario was to process meat from exclusively domestic livestock.215 Given this incentive to use exclusively US-origin livestock, we see evidence for a "decreased US demand" for Canadian and Mexican imported livestock that would be reflected in a widened price basis caused by the COOL measure.216 In light of this, it is not clear that the "decreased US demand" referred to by the United States in these proceedings would bias an interpretation of a widened price basis as a decrease in the export price of livestock.
5.58.
The United States further suggested at the meeting with the Arbitrator that added regulatory compliance costs associated with the COOL measure could be expected to lead to price increases for US livestock. In the case of the COOL measure, we understand such costs to refer to modifications of production facilities, labelling capacities, and other fixed costs, as outlined in the regulatory impact analysis of the 2009 Final Rule.217 Even assuming that it were shown that such costs increased the price of livestock, we note that costs of this nature would be non-discriminatory in that they would be incurred (and potentially passed upstream to livestock producers) independently of the particular origin of the livestock used.218 In a price basis analysis, such non-discriminatory costs would not necessarily have any impact on the price basis and, thus, would not result in any overestimation of the COOL measure's negative impact on export prices.
5.59.
With regard to the effect of decreased Canadian and Mexican exports of livestock to the United States, we recall that an important feature of the North American livestock market is the relative size of US production and demand in relation to the livestock exports of Canada and Mexico.219 It is uncontested in these proceedings that the share of livestock imports within the US market remains small and within the ranges reported in earlier phases of these disputes.220 For example, the total US livestock slaughter in the baseline year of 2014 was 30,859 million head of cattle.221 In the same year, imports of Canadian feeder and fed cattle comprised approximately 3 per cent of total US cattle slaughter222, and Mexican feeder cattle comprised approximately 4 per cent of total US cattle slaughter.223 Imports of Canadian feeder pigs and fed hogs comprised approximately 5 per cent of the total US slaughter of 106,879 million head of hogs.224 The small share of Canadian and Mexican imports compared to total US slaughter is consistent with the premise that "[t]he dominant factors in the US market are conditions that surround livestock of US origin".225 The small import share is also consistent with nearly perfect import demand elasticities within the United States226, with the result that exporters of livestock to the United States are "price-takers" according to US import demand. It follows from high import demand elasticities that any decrease in import quantities would result only in a very small (if any) increase in the prices set according to conditions within the US market.227
5.60.
Thus, while it is theoretically possible that a trade-restrictive measure could lead to higher prices in the importing country in addition to lower prices in the exporting country, we do not find compelling evidence that this has been the case with respect to the COOL measure in the US livestock market. Indeed, we note that there are indications that the COOL measure may have even led to decreased livestock prices within the United States. In a study on the changes in economic welfare of US consumers, producers, processors, and retailers resulting from the implementation of the (original and amended) COOL measure, the results of a multiple-sector EDM for beef, pork, and poultry sectors conclude that the COOL measure reduced the US price of feeder cattle, as well as slaughter (fed) cattle and hogs.228
5.61.
Finally, we note the parties' agreement that there are a series of statistical problems that arise in attempting to econometrically estimate the impact of the COOL measure on actual prices.229 Canada and Mexico explain that such a specification would require inclusion of a large number of explanatory variables, for many of which data are not available.230 Additionally, they acknowledge that such a specification faces certain statistical issues. In particular, they submit that some explanatory variables suffer from unit root problems (i.e. they are non-stationary)231 or may be endogenous variables (i.e. those that are themselves impacted by the dependent variable).232 In light of this, our conclusion is not altered by the fact that specifications using absolute prices indicate that the COOL measure increased US prices. The United States characterizes as "mis-specified" the very same price model it uses to yield such results.233 The upward trend of livestock prices underscores the need for a methodology that is capable of isolating the negative effect of the COOL measures amidst the multiplicity of factors that may have contributed to overall price increases. As demonstrated by the United States itself, an econometric analysis of absolute price levels is inadequate for this purpose.
5.62.
In sum, we consider that the COOL measure's impact on the price basis is an appropriate measure of its impact on Canadian and Mexican export prices. We note that the object of Article 22.6 proceedings is to ascertain the level of nullification or impairment, and Canada's and Mexico's use of price basis is suitable for this purpose under the specific circumstances of this case. The COOL measure is a factor that, as found in prior stages of these disputes, differentially impacts the competitive opportunities (and prices) of livestock from different origins as compared to domestic livestock. A price basis regression effectively controls for other factors that would have had the same effect on North American livestock prices, and is therefore apt to identify and isolate the impact of the COOL measure. In the next section, we discuss the application of this logic to specific variables and the rationale for their inclusion or omission in a price basis model.

5.2.1.2 Variable omission

5.2.1.2.1 Arguments of the parties

5.63.
Canada and Mexico each control for a limited set of variables in their respective model specifications for the estimation of the impact on price basis.
5.64.
The United States argues that both Canada's and Mexico's model specifications suffer from variable omission by failing to include a number of factors affecting the North American livestock and meat markets during the time-period reviewed.234 According to the United States, the estimations of the impact of the original and amended COOL measure account not only for the original and amended COOL measure's own effects but also capture some impacts of the missing variables.235 The United States submits that "it is important to ensure that, in determining the level of nullification or impairment, trade effects attributable to a factor other than the measure at issue are not attributed to the measure at issue since that would result in an erroneous level of nullification or impairment."236
5.65.
The United States thus contends that the econometric analysis presented is "insufficient to isolate the effects of the amended COOL measure" and that, "[t]o be robust, this methodology must systematically account for all relevant supply and demand shifters".237 The United States lists a number of independent variables that should be controlled for, which "include, but are not limited to": economic fluctuations and recession; long-term unemployment; increased feed costs; shifts in Canadian and Mexican livestock and meat processing; shifting transportation costs; weather patterns and drought; impacts of animal disease such as bovine spongiform encephalopathy (BSE) in the Canadian herd; and increased demand for meat during US holidays.238
5.66.
Canada and Mexico both reject the United States' contention regarding variable omission, arguing that their specifications include all the relevant exogenous variables to measure the causal impact of the COOL measure on prices.239
5.67.
Canada submits that the inclusion of variables in the model should be based on objective criteria, namely: (1) economic reasons to believe the variables have a causal impact; (2) the variables must be "clearly exogenous"; and (3) the variables must not be "temporally correlated with the dependent variable in some non-causal or random way to avoid biasing impacts of other variables."240 Additionally, Canada contests the need to include variables that do not differentially impact livestock prices and for which excluding the variable from the model does not lead to systematic bias in the estimates of interest, i.e. the impact of the COOL measure.241
5.68.
Mexico notes that "only exogenous variables that have a causal impact should be included as explanatory variables … [that] there will be omitted variable bias only if the omitted variable is correlated with the variable of interest … [and that] the United States has failed to explain why any of the 'omitted variables' it has identified … would have a differential impact on the price of imported Mexican cattle."242 Furthermore, Mexico argues that their "Methodology Paper uses a careful approach to include only the variables that are economically relevant in the regression models."243 Mexico points out that because they use the price of Mexican feeder cattle in New Mexico and Texas, and compare that to the price of United States' feeder cattle at the same locations, only a limited number of factors can explain any difference.244

5.2.1.2.2 Analysis by the Arbitrator

5.69.
In an econometric regression with price basis as the dependent variable, the relevant explanatory variables are those that affect the difference between export prices and US prices of livestock. In other words, only variables that have a differential impact on livestock prices need to be controlled for and included in the model specification. As described above245, variables having an equal impact on export and US livestock prices will have no impact on the price basis, and therefore would not need to be included in the model.
5.70.
In their respective methodology papers, Canada and Mexico have each controlled for certain variables in their model specifications. The control variables included by Canada and Mexico correspond to their differing definitions of price basis – while Canada compares the price of Canadian livestock in Canada with the price of US-origin livestock in the United States, Mexico compares the price within the United States of both Mexican-origin and US-origin livestock. To address the issue of variable omission, we briefly outline the price basis specifications used by Canada and Mexico to discuss the variables they have each controlled for in their proposed methodologies. We then turn to the question of additional relevant variables that the United States contends have been omitted.
5.71.
Canada and Mexico use similarly specified equations to estimate the impact of the COOL measure on the price basis. Each price basis specification includes parameters representing the effects of the original and amended COOL measures. The price basis specifications also include a "lagged dependent variable" to measure the relationship between the price basis at a certain point in time t and the price basis in the previous period t-1 (i.e. the previous week for Canada and the previous month for Mexico). The goal of the equations is to estimate the magnitudes of the parameters for the original and amended COOL measures, as well as the lagged dependent variable, in order to calculate the change in price (∆P) in the determination of export revenue losses caused by the COOL measure.
5.72.
This can be illustrated in the following equation used by Canada for its price basis specification246:

Pet - Pust = α + βZt + γ1(DCOOL1) + γ2(DCOOL2) + δ(Pet-1 - Pust-1) + vpt

In this equation,Pet represents the Canadian export price in period t, and Pust represents the price in the United States; the difference between these two prices (Pet - Pust) is the price basis as defined by Canada. Mexico uses a similarly specified equation with respect to its definition of price basis, namely the difference between prices of Mexican-origin and US-origin feeder cattle within the United States. Both Canada and Mexico assign parameters for the original and amended COOL measures. In the above equation, parameters γ1 and γ2 represent the effects of the variables for the original (DCOOL1) and amended (DCOOL2) measures. Parameter δ corresponds to the "lagged dependent variable" (Pet-1 - Pust-1), which reflects the tendency for causal factors to have impacts that linger for more than one period, and measures the degree to which impacts gradually dissipate over time.

5.73.
The other parameters in the above price basis specification are intended to capture the impact of other explanatory variables with respect to price basis. Parameter α is the intercept of the equation, and the random error term vpt represents random events or drivers that have impacts on exports but that are not accounted for by, and are uncorrelated with, the included explanatory variables.
5.74.
The parameter vector β represents the effect of control variables in the set Zt on the prices. In Canada's price basis specification, this includes: (a) monthly dummy variables to represent seasonality in the variation of cattle and hog export prices; (b) changes in the exchange rate between U.S. dollars and Canadian dollars; (c) dummy variables for two BSE-related events specific to Canadian cattle; and (d) a dummy variable for changes in hog processing capacity resulting from closure of a plant.247 In Mexico's price basis specification, this includes: (a) monthly dummy variables to represent seasonality in the variation of cattle and hog export prices; and (b) a dummy variable for drought events.248
5.75.
Both Canada and Mexico have thus controlled for certain variables in the benchmark specifications presented in their methodology papers. The United States contends that relevant explanatory variables have been omitted from these specifications, and that "all other relevant explanatory variables must be controlled for to isolate the impact of COOL on the dependent variable".249
5.76.
We agree that omission of a relevant explanatory variable could potentially bias the estimates of the COOL impact, specifically by erroneously attributing effects to the COOL measure that are actually caused by other factors.250 However, the determination of whether a relevant explanatory variable has been omitted turns on the specific nature of the dependent variable in relation to the allegedly omitted variables in question.
5.77.
In this connection, the parties agree on certain general criteria for assessing whether a given variable should be included in econometric modelling.251 First, including the variable must be consistent with economic theory and logic in terms of that variable's impact on the dependent variable. In the absence of an economic rationale justifying inclusion of a variable, the omission of that variable does not amount to misspecification of the model. Second, the variable must satisfy various econometric conditions such as being exogenous in the sense that it is not itself caused or impacted by the dependent variable.252 Other econometric conditions include correlation with other variables in the model, including other explanatory variables and the error term. Finally, a relevant consideration is the data that are available for a variable, and the extent to which a proxy variable may introduce measurement errors in the regression analysis.
5.78.
We apply these general criteria to the price basis specifications of Canada and Mexico outlined above. As a threshold consideration of economic theory and logic, we recall that we have accepted the use of price basis as the dependent variable, which means that only variables that would be expected to have a differential impact on prices should be included. The United States asserts relevant omitted variables include, but are not limited to, a wide variety of factors relating to the livestock production process, macroeconomic trends, market participants' behaviour, and animal disease.253 While it is not contested that such factors could be important to an assessment of absolute export prices, the fundamental question is whether any given factor would affect the prices of Canadian and Mexican livestock differently from those of US livestock.
5.79.
We recall that the products under considerations are "like products" in the sense that they are distinguished solely on the basis of their origin, rather than any of the well-established criteria for likeness.254 Given the integrated nature of the North American livestock market and the arbitrage conditions equalizing price differences, the key variables of relevance to our analysis are those representing trade barriers and differential transaction costs (such as transport costs). Other than the impact of such variables on the price basis, the market conditions in which the products in question are traded suggest that products distinguished only by origin would otherwise sell for the same price in the same place. In our view, therefore, the price basis specification does not need to include all potential supply and demand shifters within the relevant markets, as contended by the United States, but only those that account for trade barriers and transaction costs.
5.80.
We note that the logic of "differential transaction costs" applies differently to the price basis specifications of Canada and Mexico due to the different ways in which they have defined price basis. In Mexico's case, the price basis is a comparison of prices of Mexican and US-origin livestock in the same place, i.e. within the United States.255 As explained by Mexico, "[t]he price paid to Mexican producers is slightly lower than the price paid for Mexican feeder cattle in the United States because of transaction costs (including transportation)".256 Transaction costs stemming from transportation and exchange rates "have already been incurred" at the time of sale in the United States, and "the price of Mexican feeder cattle is determined solely by the valuation for feeder cattle by U.S. buyers".257
5.81.
This is distinct from the comparison made by Canada between livestock prices in Canada and in the United States. The price of livestock within Canada does not similarly reflect additional costs associated with transport to the United States and exchange rates between Canada and the United States. We note that Canada has controlled for changes in the exchange rate to account for the fact that "short term movements in the exchange rate may affect prices" at which exported livestock are sold.258 However, Canada has not controlled for transportation costs of exporting cattle to the United States. Such costs are a basic factor of trade costs that could account for a difference between the price of products that are identical apart from origin.259 Accordingly, we consider that such transport costs should be controlled for in a price basis regression in which price basis is defined in the manner proposed by Canada.260
5.82.
Apart from the inclusion of transport costs in Canada's price basis specification, we accept the explanatory variables that Canada and Mexico have each controlled for in their respective models. With regard to additional variables that the United States contends have been omitted, we decided to focus on a limited number of variables, based on the criteria above, that may have impacted the price basis, namely: (a) economic recession; (b) other competing imports from Canada and Mexico; (c) feed costs; and (d) drought. It is not contested that these variables may have some differential impact on the supply and demand curves within Canada, Mexico, and the United States. However, under full adjustment to arbitrage conditions in a highly integrated market in which exporters are price-takers, we do not consider that these variables would affect the price gap between imported and domestic products in the US market. For example, despite the different timing and severity of economic recession in the United States261, the price basis would theoretically remain unchanged if arbitrage occurred without frictions. Likewise, factors impacting the supply of livestock from Canada and Mexico would be of limited relevance to a price basis analysis where prices are set according to conditions within the US market, and where arbitrage by producers and buyers operates to equalize temporary price differences.
5.83.
Nevertheless, adjustments to price levels through arbitrage are not instantaneous, and time lags in the adjustment process can theoretically account for temporary changes in the price basis. In order to control for these impacts, we have reviewed empirical evidence on the implications of including the above variables in the price basis specifications of Canada and Mexico, particularly the extensive material provided by the parties in response to written questions from the Arbitrator.262 This includes the results of regressions including these variables separately and all together in a single specification, with results from different proxies for each variable, namely: (a) dummy variables and unemployment rates for economic recession; (b) data on other competing imports from Canada and Mexico; (c) corn and barley prices as well as future prices for feed costs; and (d) drought monitor reports as well as a dummy variable for drought.
5.84.
We examined the results of including these variables in levels and in first differences (i.e. the difference between the value of a variable at a given time t and the preceding time t-1). The reason for including certain variables in first differences was to capture the fact that the change in such variables (e.g. proxies for recession) is the shock that is the relevant potential impact on price basis, rather than the level of such variables. In addition, running the regression with certain variables in first differences corrected for unit roots.
5.85.
Based on our review, we do not find compelling empirical evidence for the inclusion of the additional variables in the price basis specifications of Canada and Mexico. These variables (and respective proxies), either included separately or all together, are not always and consistently statistically significant in any discernible pattern across the different animal categories (i.e. fed/feeder cattle/hogs) when either a 5 or 10 per cent significance level is used as the criterion for the level of significance.263 Thus, there is no consistent empirical evidence that any of the variables examined, or all of them together, is an explanatory factor that should be included in the price basis specification. Furthermore, we note that the estimations of the impact of the COOL measure on the price basis are robust to inclusion of these additional variables that the United States argues have been omitted. In other words, the COOL measure parameters, which are the primary interest of this analysis, remain consistent and statistically significant even when these additional variables are included.
5.86.
Accordingly, we consider that Canada's proposed price basis specification omits transport costs, while Mexico's price basis specification does not suffer from omission of this variable. As regards other variables discussed in this section, we do not find conclusive evidence that they need to be included in a price basis estimation. However, to the extent statistically significant estimates of such variables (though inconsistent across animal categories, the proxy used, and whether estimated in levels or first differences) confirm the fact that adjustment lags and market frictions may lead to such variables affecting the price basis, we reserve consideration of such variables for the purposes of checking the robustness of our own determination of the COOL measure's impact on export prices.

5.2.1.3 Price impact on Canadian feeder pigs

5.2.1.3.1 Arguments of the parties

5.87.
As described above in section 3.1, Canada does not use regression analysis to estimate the price impact on feeder pigs. Canada explains that "it was not possible to estimate these impacts statistically" because "no consistent time series of price data amenable for statistical analysis is available for feeder pigs in Canada."264 Canada relies instead on a descriptive comparative analysis of price information based on [[BCI]] invoices provided by a Canadian firm trading in feeder pigs on either side of the border.265 Canada also relies on witness statements submitted by the president of that firm. In the witness statements, it is stated that the firm controls [[BCI]] per cent of the export market of feeder pigs to the United States.266
5.88.
Canada pairs the invoices into [[BCI]] pairs of transactions, each pair including a cross-border (Canada-US) and an intra-US (US-US) transaction to allow for comparison.267 The invoices cover weanlings (i.e. baby pigs weighing less than 7kg) and larger feeder pigs.268 They relate to a period between July 2012 and early 2015, thus comparing the price difference between US and Canadian feeder pigs before and after the entry into force of the amended COOL measure.269 The discount effect established on the basis of this comparison is then added to the discount effect of the original COOL measure, which is derived from witness statements, to form a total discount suffered as a result of the COOL measure.270 The respective average discounts for the two weight categories (i.e. weanlings and larger feeder pigs) are weighted by 70 and 30 percent, representing, according to the witness statement, Canada's respective export shares for weanlings and larger feeder pigs.
5.89.
The United States notes that the data are taken from a single firm in the Canadian market, which provided only a limited set of invoices.271 The United States further notes that the information from the invoices does not indicate (i) the location of the purchasers, (ii) the size of the pigs, or (iii) the volume of pigs sold per transaction.272 The United States argues that "there is no way to tell if volume discounts, transportation costs, or different product specifications play a role in the alleged price basis."273 According to the United States, it is also likely that the firm submitting such data participated in more transactions than evidenced by the limited number of invoices submitted, and any long-term trend should be discerned from the entirety of transactions rather than a small sample.274 The United States takes the view that Canada should have submitted the entire sales files of the firm in question275, and further submits that Canada should have submitted sales files from more than one "large" feeder pig provider.276 The United States also suggests that a consistent time series of monthly data is available in the form of Agriculture and Agri-Food Canada reports as well as US Census Bureau data.277
5.90.
Regarding the reliability and representativeness of the invoice data, Canada notes that the evidence submitted is consistent with the price trends determined through data for the other three categories of livestock.278 Additionally, Canada states that the [[BCI]] paired transactions were chosen "to ensure that the invoices represented average volumes and average transportation costs."279 Canada further notes that witness statement submitted is a sworn statement.280 Canada submits that it has used the "best available information", for its calculations of price impact on feeder pigs, and that a lack of government data should not deny Canada losses for the drop in feeder pig prices resulting from the COOL measure.281 Following a question from the Arbitrator, Canada confirms that it does not have access to transactional information via other governmental sources.282

5.2.1.3.2 Analysis by the Arbitrator

5.91.
We begin by observing that Canada is free to decide which approach to adopt in order to estimate the price impacts of the COOL measure on its trade in feeder pigs. In particular, the fact that it uses an econometric approach for all other price estimations does not mean that it has to do so for feeder pigs as well. The United States does not contest this, nor does the United States challenge the comparative analysis of invoices as an invalid approach per se.
5.92.
What the United States challenges is the way in which the comparative analysis has been carried out, and in particular the representativeness and comparability of the data used.
5.93.
We asked Canada for additional invoices and additional information on certain issues. We also asked Canada to econometrically estimate the price impact using available monthly data from the US Census Bureau or AMS in order to compare the outcome. We acknowledge that Canada has made every effort to fully answer our requests and questions within the short timeframes applicable in these proceedings. The replies of Canada, however, do not sufficiently address the concerns that the United States has raised, and indeed raise further questions about the data used to estimate the impact on the price of feeder pigs.
5.94.
We have now a total of [[BCI]] invoices submitted, which represents 3.2 per cent of the total number of [[BCI]] transactions during the period of 2012 to 2015 (namely [[BCI]]). We note the United States' view that Canada should have submitted all invoices of [[BCI]] as well as invoices from other Canadian traders. Apart from the broader evidentiary issue this may raise, we see a number of factors that call into question the representativeness and comparability of the data actually submitted.
5.95.
First, we note that no invoices on weanlings from the pre-amended COOL period seem to be available.283 This raises questions given that (1) [[BCI]]284 and (2) we are asked to rely on [[BCI]] witness statement regarding the applicable discount on weanlings during that same period. No further explanation has been provided in this regard.
5.96.
Second, we note that there are significant differences in the level of discount (i.e. price differential) identified between the witness statement submitted in October 2013 and the first set of invoices and, again, between the first set of invoices and the additional invoices submitted at our request. The witness statement identified the average discount for larger feeder pigs for the amended COOL measure to be between USD 5 and USD 10.285 The first set of invoices identifies the same discount as USD 10.88, whereas in the additional set of invoices, that discount increases to USD 14.97.286 In addition, the decline in discount between the original COOL measure and the amended COOL measure, as identified in the witness statement, is a median of USD 3.50.287 The same differential between the original COOL measure and the amended COOL measure is USD 4.87 in the first set of invoices and USD 10.63 with the additional set of invoices (i.e. more than twice as large).288 With respect to weanlings, the average discount for the amended COOL measure goes from USD 9.18 in the first set of invoices to USD 10.78 in the additional set of invoices.289 Canada submits that this demonstrates that the original invoice sample included is conservative.290 In our view, however, these significant differences call into question the representativeness and reliability of the "discount" identified, in particular with respect to larger (non-weanling) feeder pigs.
5.97.
Third, even assuming that the Canadian prices submitted are representative of the market price, we have doubts about the representativeness of the US prices in the invoices submitted by [[BCI]]. Canada states that no more than [[BCI]] per cent of [[BCI]] transactions are US‑US transactions. In other words, out of the total of [[BCI]] transactions that [[BCI]] carried out during the period of July 2012 to May 2015, only a maximum of [[BCI]] transactions were US‑US transactions. A sample comparison of US transaction prices indicated in the invoices with AMS monthly average rates of US prices for the same dates shows that the average invoice price of US feeder pigs is considerably higher. For example, the greatest price differential in the invoices submitted is for September 2014. The US‑US transaction for that month indicates an average rate for a 40lb pig of [[BCI]]; the AMS monthly average rate for the same month is USD 80.8.291 We note that the AMS price for September 2014 [[BCI]].292 This shows that use of invoice data can lead to significantly greater price differentials than what is indicated by average monthly US prices reported by the AMS. In sum, we are not convinced that US prices submitted by [[BCI]] are actually representative of the US market given their relatively high price volatility.
5.98.
Fourth, the varying weight of feeder pigs is another reason to call into question the comparability of the price data submitted. In some of the paired invoices, the weight of the pigs differs substantially.293 We understand that the price for feeder pigs is calculated on the basis of weight and that a "rate" is applied that differs from the first 40lb to the next 20lb, and again to any additional weight over 60lb.294 In some paired invoices, the weight of the pigs differs between 10 and 20kg. How that weight difference affects the final price, and thus the price differential, depends on the rate that is applied. In one paired sample submitted, the rate applied to Canadian and to US pigs differs by only [[BCI]], whereas the average differential of the actual price paid shows [[BCI]] for Canadian pigs. Based on our review of the evidence, we understand the reason for this differential to be that the Canadian pig is 13kg (about 26 per cent) heavier than its US counterpart and, therefore, is sold at a higher price. In another sample, the price rates differ by [[BCI]], but the price differential is only [[BCI]]. Our understanding is that this is because the Canadian pig (sold at the cheaper rate) is 15kg (about 25 per cent) heavier, which increases its final price and therefore makes the price differential comparatively smaller. In sum, different weights have an impact on the price differential. Therefore, where weights differ significantly, it calls into question the comparability of the prices. This is particularly important because Canada considers that the price difference of comparable Canadian and US feeder pigs is entirely due to the COOL measure.
5.99.
Finally, we note some lacunae in the invoices submitted. First, it was stated that no invoices were available for the month of January 2014295, and therefore there are no data during one month in the time series. Furthermore, we note invoice No. 71, as listed in the table, has not been submitted. Instead, we find an invoice to a different company with a price listed as [[BCI]] per weanling instead of the [[BCI]] listed in the table. These irregularities further call into question the statistical reliability of the invoices Canada submitted to estimate the impact of the COOL measure on feeder pig prices.
5.100.
As noted above, we had asked Canada to submit an econometric estimation of the feeder pig price based on available monthly data. The purpose was to compare the result to the results of the comparative analysis of the invoices to get a sense of how far apart these two approaches are. However, we are unable to compare the two results because the two approaches rely on data for different weight categories. While the monthly data used in the econometrics cover weanlings below 7kg and feeder pigs between 7kg and 23kg, the invoice data also include larger feeder pigs between 23kg and 50kg.296
5.101.
Canada contends that it has used the best available information for its calculations of price impact on feeder pigs, and that a lack of government data should not deny Canada losses for the drop in feeder pig prices resulting from the COOL measure.297 We agree that if and where a Member has submitted the best available information, it might be appropriate for an arbitrator to decide to accept that information in that particular proceeding. However, we note that in this arbitration, alternative data sets have been proposed that do not suffer from the fundamental issues of representativeness and comparability described above.298 Thus, for the reasons set out above, we are not convinced that the information submitted by Canada does indeed qualify as the "best available information".
5.102.
In conclusion, we find that Canada's estimation of the feeder pig price is not sufficiently reliable. Given this finding, we do not need to address the broader question, raised by the United States, of whether Canada would have had to submit all invoices of the firm in question and also invoices from other companies.

5.2.2 Quantity impact estimation

5.2.2.1 Canada's econometric estimation

5.2.2.1.1 Arguments of the parties

5.103.
Canada uses an econometric analysis to estimate the impact of the COOL measure on the quantity of livestock exported to the United States.
5.104.
The United States argues that "[f]or the estimates in Canada's model to reflect any degree of accuracy, the variables that may have an effect on price or quantity must be accurately estimated and properly specified."299 Thus, the United States makes the same arguments against Canada's econometric determination of quantity impact that it makes in respect of the econometric determination of price basis (i.e. regarding variable omission and the suitability of econometric analysis for isolating the impact of the COOL measure).300
5.105.
Canada argues that its econometric estimations are focused "solely" on the impact of the COOL measure, and deliberately exclude "extraneous variables that would introduce concerns that would bias the measured impacts of the amended COOL measure".301 Canada submits that, "[a]s with the price basis, export quantity, by definition, reflects the difference in economic conditions between the markets in the two countries."302 Canada further states that, "[j]ust as with the price basis, variations in export quantity are driven by variations in the functions representing both the demand for imports and supply of exports."303

5.2.2.1.2 Analysis by the Arbitrator

5.106.
As described above, an accurate econometric analysis requires capturing and accounting for relevant factors (independent explanatory variables) that have an effect on the dependent variable (in this case the change in quantities exported). If accurately specified, an econometric analysis should isolate the effects of a particular independent variable (in this case the COOL measure) to determine its impact on quantities of Canadian livestock exported to the United States.
5.107.
The quantity estimation submitted by Canada, unlike the price basis specification discussed above, is in terms of absolute quantity levels. We recall that the price basis specification was appropriate given the nature of the North American livestock market and evidence of price-arbitrage conditions. Focusing on the price basis enabled exclusion of variables that did not differentially impact prices of different-origin livestock. As noted by Canada in that context, "including the U.S. price in the basis specification allows one to capture parsimoniously the impacts of a host of variables that may affect livestock prices in both countries in a similar way".304
5.108.
Estimating quantity impacts in absolute levels, by contrast, does not similarly permit omission of variables that had a common impact on North American markets. Rather, such an estimation would need to account for any supply or demand factor affecting export quantities. Canada refers to "the amended COOL measure's impact on export quantity" also being "direct, i.e. it caused a loss of export shipments between Canadian and U.S. locations."305 While it is expected that the COOL measure, as the relevant explanatory variable of interest, would affect export quantities, failure to control for other such variables creates the potential for introducing bias in the COOL parameter estimates.
5.109.
We note Canada's acknowledgement that the purpose of its regression specification is "to estimate the magnitude of the impact on actual exports" that should "generate estimates that directly enter the calculation of losses".306 At the same time, Canada refers to the exclusion of factors that affect livestock export quantities, citing lack of available data and likely underestimation of the COOL impact as a result of omitting the variables.307 However, these are among many other potential variables that have been omitted in Canada's quantity specification. As pointed out by the United States, separate estimations of absolute price and quantity impacts would need to control for the same exogenous factors as explanatory variables.
5.110.
Therefore, Canada's econometric estimation does not adequately control for relevant explanatory variables that, in addition to the COOL measure, may affect livestock export quantities. Although Canada's quantity specification mirrors its price basis specification in respect of the set of control variables, the methodological advantages afforded by focusing on price differentials do not apply to an estimation of absolute quantity levels. As a result, we are unable to accept Canada's econometric estimation of the impact of the COOL measure on export quantities.

5.2.2.2 Mexico's elasticity-based simulation of quantity effects

5.2.2.2.1 Arguments of the parties

5.111.
To estimate the impact of the COOL measure on the quantity of Mexican feeder cattle exports, Mexico relies on a simulation using the COOL impact on export prices (estimated econometrically using the price basis) and a derived elasticity of export supply.
5.112.
The United States argues that this approach is "insufficient to account for the complexity of the feeder cattle market in Mexico and the United States, much less to account for linkages to demand for fed cattle and beef or to substitute products such as pork."308 The United States is of the view that the calculation should account for all factors influencing quantity outcomes, including supply and demand effects in the United States and in Mexico, as well as the impact of exports from Canada to the United States.309 The United States suggests that "Mexico’s estimation of the quantity impact is based on a formula which assumes 100 percent pass through of the bias inherent in the price basis estimate into the quantity change simulation. The result is an estimate that grossly overestimates the effect of COOL on U.S. imports of feeder cattle from Mexico."310
5.113.
Additionally, the United States argues that the value of the export supply elasticity used to simulate the counterfactual change in export volumes is unproven, un-reviewed, and derived with little explanation.311 According to the United States, Mexico's elasticity of export supply is based on a single year, 2012, a period of time most certainly affected by drought and other factors.312 The value of the export supply elasticity also appears to make unsupported assumptions about the rate of export.313
5.114.
Mexico notes that its equation for determining the elasticity of the export supply curve is identical to an equation used in the United States' EDM.314 Additionally, Mexico submits that this single equation is "sufficient and does not need to account for the complexity of the feeder cattle market in Mexico and the United States because, as explained previously, this is accounted for in the … price basis regression."315 Mexico further notes that exports of livestock from Mexico and Canada "represent a small share of the total U.S. livestock market."316 Mexico therefore asserts that "[c]hanges in export volumes from Mexico and Canada would thus have a small impact on U.S. livestock prices."317 According to Mexico, the geographic size of the United States also limits direct competition between Mexican and Canadian cattle.318
5.115.
Regarding its estimated export supply elasticity figure, Mexico contends that the figure was "derived based on observed data in a transparent way."319 Mexico notes that the derived elasticity value of 4.0 is reasonable given the size and structure of Mexico's cattle market, as well as empirical evidence on supply and demand elasticities.320 Mexico asserts that 4.0 is a "conservative estimate" given the empirical evidence and the "length of run over which the market adjusts to the introduction or the removal of COOL measures."321 Mexico further argues that "[t]he calculation of the export supply elasticity builds on values from the literature for Mexico’s domestic demand and supply of feeder cattle and uses the observed share of feeder cattle produced that it exported to the United States. This is the proper method to calculate the elasticity."322

5.2.2.2.2 Analysis by the Arbitrator

5.116.
In contrast to Canada's econometric estimation of quantity impact, Mexico relies on an elasticity-based simulation. Based on the econometrically estimated impact of the COOL measure on export prices, Mexico simulates the corresponding impact on export volumes using a derived elasticity of export supply.
5.117.
Elasticity measures the responsiveness of one economic variable to a change in another variable. Mexico derives the export supply elasticity value by using a rearranged expression of the elasticity of the export supply curve. The elasticity of the export supply curve measures how the volume/quantity of exports responds to changes in the export price. In mathematical terms, the price elasticity is defined as the ratio between the percentage change in export quantities and the percentage change in export price in relative terms.
5.118.
The counterfactual change in export quantity is expressed as the product of: (a) the counterfactual change in export price; (b) the inverse of the export price in the baseline period; (c) the export quantity in the baseline period; and (d) the elasticity of Mexico's export supply.323 This can be expressed using the following equation where Qc and Pc are the observed quantity and export price of feeder cattle, ΔP is the estimated impact of the COOL measure on the export price of feeder cattle, and εe is the export supply elasticity324:

[SEE IMAGE IN SOURCE DOCUMENT]

5.119.
While data on export price and export quantity in the baseline year are available, and the counterfactual change in export price was estimated econometrically, the elasticity of Mexico's export supply is the only parameter that remains to be determined.
5.120.
Because the value for the elasticity of export supply is not directly available in the literature and is technically difficult to estimate empirically325, Mexico derives its export supply elasticity of feeder cattle as a function of three variables: its own domestic supply and demand elasticities for feeder cattle and the share of exports of feeder cattle of Mexican supply. In mathematical terms, this formula can be expressed as follows where ω is the share of exports, εs is the elasticity of domestic supply, and η is the elasticity of domestic demand326:

[SEE IMAGE IN SOURCE DOCUMENT]

5.121.
We note that this is a well-established formula in the economic literature used for the calculation of export supply elasticities.327 Ideally, using this formula would simply require inputting the relevant figures for Mexico's domestic supply and demand elasticities with a straightforward computation of its export share.
5.122.
With regard supply and demand elasticities for feeder cattle, direct estimates of these elasticities are not available for Mexico. For this reason, Mexico uses estimates of long-run supply and demand elasticities for US feeder cattle. Mexico explains that "the supply elasticity of feeder cattle is determined by biological factors such as gestation period, and these biological factors are the same in the United States and Mexico".328 Furthermore, Mexico argues that recent modernization of Mexican cattle industry and other factors "make[] the Mexican cattle industry more like that of the United States".329 Thus, in Mexico's view, absent direct estimates for Mexican demand and supply elasticity, estimates for US demand and supply elasticities are reliable values of Mexican feeder cattle elasticities.
5.123.
We note that the United States has not objected to the use of US demand and supply elasticities as a proxy for Mexican demand and supply elasticity values. Furthermore, the United States also considered estimates of US-based elasticities as proxies for Mexico's elasticity in its proposed methodology (i.e. EDM) to estimate the level of nullification or impairment.
5.124.
Turning to the figure for the export share of feeder cattle (ω in the export supply elasticity formula), Mexico acknowledges that calculating the export shares of feeder cattle requires knowing Mexico's annual production of feeder cattle and exports to the United States. For this purpose, Mexico uses the figure of 1.11 million head of feeder cattle exports annually to the United States330 and calculates an "annual beef calf crop in Mexico of 4.8 million heads".331 These two figures are the basis for, respectively, the numerator and the denominator in Mexico's calculation of export share. Thus, Mexico explains that, "[a]ssuming that all feeder cattle in Mexico can be exported to the United States, with an annual crop of 4.8 million heads and exports of 1.11 million heads, yields ωe = 0.23."332 We note that the inputs for this calculation are themselves derived on the basis of various assumptions. In particular, Mexico relies on the assumption that the number of feeder cattle born and exported is the same in two consecutive years to estimate the annual beef calf crop production in 2012.
5.125.
Apart from these considerations for deriving the necessary inputs, a fundamental assumption in Mexico's calculation of export shares is that "not all feeder cattle in Mexico are eligible for export".333 According to Mexico, this is due to identification requirements, protocols to prevent importation of diseased animals, and quality considerations relating to different breeds of cattle in Mexico. In this regard, Mexico considers that value of ω = 0.75 "is reasonable given the description of the Mexican cattle industry provided in" studies of beef and cattle production in Mexico.334 Mexico then calculates from this asserted export share of 0.75 that only 31 per cent of annual calf production (i.e. 31 per cent of the 4.8 million heads derived from the assumptions described above) is eligible for export to the United States (where 0.75 = 1.11/4.8*0.31). Thus, the share of 31 per cent of production that is eligible for exportation is simply assumed on the basis of what Mexico contends is a "reasonable" value for the export share of feeder cattle.
5.126.
With regard to export share and export supply elasticity values submitted by Mexico, the United States argues that Mexico "appears to make unsupported assumptions about the rate of export, and ultimately with little explanation concludes that the export supply elasticity is 4This elasticity exceeds the appropriate level."335 The assumptions underlying Mexico's calculation of export shares are indeed integral to its derivation of the export supply elasticity with which it simulates the export quantity impacts of the COOL measure. At the same time, Mexico acknowledges that there are "no data that specifically describe […] the number of feeder cattle according to their breed".336 We note that the sources relied upon by Mexico for its export share assumptions do reflect that certain breeds are more commonly exported.337 However, the same sources indicate variation in this pattern and shifts from historical trading patterns that cast further doubt on the accuracy of Mexico's assertions about the share of export eligibility and, correspondingly, the actual share of those eligible cattle that are exported to the United States.338 In our view, the United States' agreement339 that some adjustment should be made to Mexico's export share does not amount to conceding that Mexico has correctly done so. Nor does it obviate the United States' contention that the elasticity value derived by Mexico exceeds the appropriate level.
5.127.
We note that the export supply elasticity varies inversely with the export share of a product. Thus, a smaller export share of a product in relation to total supply will lead to greater export supply elasticities, and thus greater impacts on export quantities as a result of changes in the export price. Even if Mexico's assumed value of 31 per cent of export eligibility were correct, the question then becomes whether the share of exported cattle from the limited number of eligible cattle is no more than 75 per cent. If more than 75 per cent of cattle eligible for export were in fact exported, using an export share value of 0.75 would inaccurately inflate the export supply elasticity and, consequently, the export quantity impact. The same is true if the assumed value of 31 per cent of export eligibility overstates the share of annual calf production that can be exported. Had Mexico assumed that, say, only 20 or 10 per cent of cattle were eligible for export, the resulting export share would be greater and thus yield a lower export supply elasticity than 4 (as well as a lower impact on export quantities). In either of these scenarios, the export supply elasticity (and export quantity impact) would be overstated based on unverifiable assumptions about Mexican cattle production and exportation.340
5.128.
Given the foregoing, we are of the view that export supply elasticity value of 4.0 submitted by Mexico is insufficiently supported by evidence for an assessment of whether Mexico's proposed level of suspension is equivalent to the level of nullification or impairment.

5.2.3 Data issues

5.129.
In this section, we address a number of data issues that the United States has raised in challenging the reliability of Canada's and Mexico's calculations.

5.2.3.1 Use of data other than US Census Bureau import statistics

5.130.
We begin with the United States' challenge concerning the use of data other than the official import statistics provided by the US Census Bureau, an entity within the US Department of Commerce.

5.2.3.1.1 Arguments of the parties

5.2.3.1.1.1 Canada's use of APHIS data

5.131.
We note that Canada uses weekly data from the Animal and Plant Health Inspection Service (APHIS) of the USDA for purposes of identifying the quantities of all four categories of livestock exported.341 Those quantities are used in the econometric estimation of the counterfactual export quantities as well as for the baseline values.
5.132.
The United States challenges the use of weekly APHIS data, arguing that only official monthly data are accurate and appropriate.342 Specifically, the United States claims that Canada relies on unofficial weekly cattle and hog import data derived from veterinary certificates collected by APHIS.343 The United States explains that APHIS statistics are unofficial and are not subject to publically released corrections or revisions.344 Furthermore, the fact that weekly data for cattle and hogs imports "is often revised and may not be reported for each week" causes overall data to be incomparable.345 In addition, the United States argues that using weekly data is inappropriate because weekly data introduce "significant 'noise'" into the dataset, affecting the econometric regressions and subsequent analysis.346 The United States suggests using only official monthly import data provided by the US Census Bureau data, which are less "noisy".347
5.133.
The United States re-ran Canada's econometric model using official US Census Bureau data, and after revising its initial iteration to fix certain methodological errors identified by Canada, the United States asserted that the use of APHIS data increases the estimate of export losses by almost double: the use of US Census Bureau data suggests that approximately 156,684 feeder cattle would have been exported per year, while the APHIS data used by Canada indicates that 306,176 feeder cattle would have been exported per year.348 The United States suggests that this demonstrates that Canada's model is not robust, because a robust model would provide similar results regardless of the use of monthly or weekly data.349
5.134.
Canada notes that APHIS data provide "four times as many data observations for export quantities than U.S. census data and therefore produces a more accurate and precise econometric calculation."350 Furthermore, Canada relied on APHIS data previously, and the original panel "concluded that the data was acceptable and reliable."351 Additionally, Canada argues that APHIS data are available "every week for many years in a consistent way" and are recorded by "an official government agency of the United States."352 Canada emphasizes that "[f]or all animal categories, the quantity shipped into the United States has no missing values for any week."353 Regarding "noise", Canada suggests that any random noise generated (if there is any) "tends to drive estimated coefficients towards zero, not make them seem larger and more significant".354
5.135.
Canada notes that the data recommended by the United States "do not allow measures of export animal prices for animals precisely comparable to those for which there are domestic prices in the United States [meaning that] these monthly data are not suitable for price basis regressions and no analyst uses them for this purpose."355 Additionally, "the data from USDA APHIS derives from actual border inspections of livestock shipments, whereas using Commerce Department data relies on reporting of total values of shipments and associated quantities. (Recall that these are reported by HTS code for items for which there is no import duty or other trade barrier other than animal health and COOL.)"356 Additionally, Canada states that US Department of Commerce data are not available for fed barrows and gilts for immediate slaughter, a relevant category of livestock357, and the US Department of Commerce data "mix fed hogs with old sows and boars shipped to the United States for slaughter and Canada is not claiming losses in respect of this trade."358 Canada also heavily criticised a number of alleged methodological errors in the United States' initial attempt to replicate Canada's econometrics utilizing US Census Bureau data.359

5.2.3.1.1.2 Mexico's use of AMS data

5.136.
We note that for purposes of its price estimation and its baseline values, Mexico uses weekly price data for Texas and New Mexico provided by the Agricultural Marketing Service (AMS) within the USDA.360 Mexico uses a monthly average price of two weight categories, 350lb and 550lb, representing the mid-point weights of two different weight categories.361
5.137.
The United States characterizes Mexico's use of AMS pricing data as inaccurate and inappropriate.362 The United States argues that such pricing data are not consistently reflective of the "types" of feeder cattle that are imported from Mexico due to heavy reliance on "auction data", which would not apply to feeder cattle from Mexico sold on the basis of "forward contracts or other pricing devices".363 For this reason, the United States is of the view that weekly AMS data are likely to overestimate the baseline prices for cattle, resulting in an inaccurate and inflated price basis.364 According to the United States, apart from the United States' own trade data, "Mexico’s official trade data demonstrates that the per unit export value is much closer to the per unit U.S. import value that the U.S. Census [Bureau] reports."365
5.138.
Mexico notes that the AMS data "offer an unbiased measure of the price paid for Mexican feeder cattle … The data provided by the AMS are appropriate for this analysis and in fact the United States used the same data source to calibrate its own EDM."366 Mexico notes that the US Census Bureau data relied on by the United States use customs value to determine prices, which is not based on transaction value "where imported items have not yet been sold at the time of importation, which is the case for Mexican cattle that are brought across the border to be sold subsequently within the United States."367 Mexico also states that the value of the cattle is irrelevant for customs purposes, since cattle are imported duty-free (under NAFTA) and even the United States' MFN duty rate "is appraised on the basis of animals' weight and not their value."368 Thus, since unit-values based on customs value are not based on actual market transactions, they do not offer an accurate value of Mexican feeder cattle exported to the United States.369 Furthermore, Mexico provides evidence that while "buyers and sellers may have verbal discussions prior to exportation … sales are finalized only after the cattle have crossed the border into the United States."370 Mexico also asserts that the "Mexican cattle industry reports that almost all export transactions" are in the form of "ventas directas".371
5.139.
Mexico also notes that: (a) AMS data and United States' census bureau data are broadly consistent up until 2011372; (b) a report published by the USDA in October 2014 demonstrates prices in New Mexico and Texas that are consistent with Mexico's pricing data373; and (c) the prices of Canadian feeder cattle for Canada's base-period (November 2013 to November 2014), as calculated by both Canada and the United States, "are much closer to the values calculated by Mexico for Mexican feeder cattle during the period 2014".374 Mexico emphasizes that the United States customs data do not reflect real transaction prices. In support of its arguments, Mexico provides the following evidence: (a) sample invoices375; (b) evidence that USDA data are based on sales after importation376; (c) satellite imagery of a border crossing ranch facility377; and (d) a ruling of a United States Tax Court stating that cattle usually spend no more than eight hours on the United States' side of the border before being collected by buyers.378

5.2.3.1.2 Analysis by the Arbitrator

5.140.
Calculating export revenues requires inputs of data on prices and quantities. In Canada's and Mexico's approaches to these calculations, data are needed not only for the "baseline" value (actual revenues in 2014), but also for the sample of observations used in the econometric estimation. The United States also needs data on baseline values for purposes of its calculations under an EDM. Thus, all calculations require price and export quantity information. The disagreement among the parties relates to the appropriate source from which that data should be taken.
5.141.
The United States' position is that Canada and Mexico must use import statistics from the US Census Bureau rather than data from the USDA, namely APHIS and AMS. The United States itself relies on the US Census Bureau import statistics with regard to Mexican and Canadian baseline values.379
5.142.
The fact that a party designates certain import statistics as the official source of trade data does not mean that Members are limited to using these particular data for purposes of calculating import figures. Nor are we as Arbitrator confined to reliance upon such statistics in carrying out our mandate. In our view, whether or not a specific source of data may be used for these purposes turns on whether the data are reliable and reflect as accurately as possible import quantities and/or import prices.

5.2.3.1.2.1 Canada's use of APHIS data

5.143.
With regard to APHIS data, we refer to our conclusion above in section 5.2.3.1 that Canada's econometric quantity estimation does not adequately control for factors other than the COOL measure that affect export quantities. Therefore, we are unable to rely on Canada's econometric estimates of the quantity impact irrespective of the data used. Under the circumstances, it is unnecessary to further examine the use of APHIS data in respect of Canada's quantity impact estimation.
5.144.
The issue that remains is whether APHIS data may be used for purposes of establishing baseline values. APHIS is an entity within the USDA and, thus, is a government source. We do not see a reason to consider data from this source to be inherently less reliable than the alternative source of US Census Bureau statistics. The question, in our view, is which data best reflect actual import quantities. We observe that the quantities identified by APHIS are generally lower380, but differ only slightly from the quantities identified by the US Census Bureau.381 Given that the differences are modest and, in addition, could be explained by inaccuracies in either source, we see no reason to consider the use of APHIS data to be any less accurate or reliable than US Census Bureau data.
5.145.
As regards slaughter hogs, we note that there is a considerable discrepancy between the quantity reported by the United States and the quantity reported by Canada.382 However, we observe that in its calculations, the United States uses only a percentage of the quantity reported.383 The reason for doing so, according to the United States, is that the relevant tariff line from which the United States derived the overall quantity does not distinguish between barrows and gilts slaughtered to obtain muscle cuts (to which labelling requirements apply), and sows and boars slaughtered for other types of meat (to which labelling requirements do not apply).384 In recognition of the broad scope of products included in source data, the United States uses only a percentage of the total for purposes of its calculations. Of interest is that the percentage is derived from APHIS data, confirming that the United States considers such data sufficiently accurate for this particular purpose.385 It yields a number that differs only slightly from the actual APHIS number (382,000 heads as opposed to the 405,124 reported by Canada).386
5.146.
Given the foregoing, we see no reason to conclude that APHIS data reports import quantities any less accurately than US Census Bureau data does. We therefore reject the United States' argument that Canada's use of APHIS data is inappropriate.

5.2.3.1.2.2 Mexico's use of AMS data

5.147.
With regard to AMS data, we note that these data are generated by an entity within the USDA, and thus come from the same government source as do the APHIS data referred to above. As we observed earlier, we do not see a reason to consider this US government source to be any less reliable than the other US government source, which is the US Census Bureau (within the Department of Commerce), from which the import statistics are sourced. We observe in this regard that the United States itself partially relies on AMS data for purposes of identifying its own domestic price.387
5.148.
The United States explains, however, that in respect of prices for Mexican cattle, AMS reporting relies on auction data, which would not apply to feeder cattle from Mexico sold on the basis of "forward contracts or other pricing devices."388
5.149.
In response to this argument, Mexico submits evidence showing that the AMS data it relied upon covers 71 per cent of the total quantities of cattle exported to the Unites States.389 Mexico also submits evidence to show that AMS prices closely correspond to sales prices in so-called "ventas directas" (direct sales).390 A witness statement submitted by Mexico describes these direct sales as sales that take place once the cattle have crossed the border.391 Further evidence submitted by Mexico demonstrates that exported cattle are picked up by buyers shortly after having crossed the border, which confirms that the prices paid are "fob" (as reported by AMS) and do not contain any US added value.392 A further witness statement testifies that almost all export transactions are done through such direct sales.393 Thus, on the basis of the evidence submitted by Mexico, it does not seem that AMS price data rely on any auction data at all; rather, they accurately reflect sales prices as agreed in the "direct sales", which are the main kind of cross-border transactions in cattle.
5.150.
The United States, on the other hand, submits that it has no information on what kinds of sales take place between Mexico and the United States.394 Furthermore, the United States is unable to show that its own "unit value" as reported in the US Census Bureau import statistics reflects actual sales prices rather than just an "entered value", which, as the United States itself points out, "is not the value of a later sale when the animal is already in the United States".395
5.151.
As noted above, any data used must reflect as accurately as possible import prices or import quantities. Mexico has submitted convincing evidence that AMS prices reflect actual import prices of Mexican cattle as accurately as possible, and in any event, more accurately than the "unit value" reported by the US Census Bureau.
5.152.
In conclusion, we reject the United States' argument that the use of AMS pricing data by Mexico is inaccurate or inappropriate.

5.2.3.2 Canada's sample period in respect of its cattle specification

5.153.
We note that, for purposes of the cattle specification in its econometric estimation of price as well as of quantity, Canada uses a sample period starting in September 2005 and continuing through January 2015.396

5.2.3.2.1 Arguments of the parties

5.154.
The United States asserts that Canada's utilization of data between 2005 and 2015 fails to accurately evaluate the impact of the COOL measure, since "the 'pre-COOL' period used is concurrent with the BSE event and its lingering effects."397 Since these and other factors ("such as the effects of additional BSE episodes") are unaccounted for by Canada, the United States contends that the model may misattribute effects to COOL that were caused by these other factors.398
5.155.
Canada suggests that by July 2005 the impact of BSE was not important as "trade in fed and feeder cattle had resumed."399 Since the original COOL measure was only implemented at the beginning of the fourth quarter of 2008, Canada considers that sufficient time had elapsed from the BSE event, "which had by that time been resolved for young cattle imports."400 Additionally, Canada argues that lingering impact for older animals is resolved through the use of dummy variables.401 At the request of the Arbitrator, Canada estimated the impact of extending the sample period to 2003. Canada noted that no export price data were available for the period due to the ban on cattle imports.402 With respect to the quantity estimation, Canada noted that for fed cattle "[t]he impact of the amended COOL measure is just slightly smaller than in the base data set."403 For feeder cattle, "[t]he COOL impact is just slightly larger than when the shorter sample is used for the regressions."404

5.2.3.2.2 Analysis by the Arbitrator

5.156.
We concur with Canada, in respect of the econometric price estimation, that it is impossible to extend the sample period to 2003. The reason is that there are no price data for the period 2003 to 2005 given that a ban was in place due to BSE.405 The question whether to extend the period to 2003 therefore only concerns the econometric quantity estimation.406 We concluded above in section 5.2.3.1 that Canada's econometric quantity estimation does not adequately control for factors other than the COOL measure that affect export quantities, and thus cannot be relied upon as specified by Canada. Therefore, there is no need to make findings on whether the sample period for such estimation should be extended to 2003.

5.2.3.3 Starting dates for COOL dummy variables

5.157.
We note that Canada, in its econometric estimation of price as well as of quantity, uses two dummy variables in its specification to account for the effect of the original and the amended COOL measures. The dummy variable for the original COOL measure has the starting date 29 September 2008. The dummy variable for the amended COOL measure has different starting dates depending on the livestock in question: 23 May 2013 for small-size feeder cattle and feeder pigs; 1 July 2013 for intermediate size feeder cattle; and 2 November 2013 for fed cattle and fed hogs.407

5.2.3.3.1 Arguments by the parties

5.158.
The United States submits that Canada did not correctly define the time-periods to create the original and amended COOL measure dummy variables.408 The United States recalls that the 2009 Final Rule "became effective on 16 March 2009", but that the dummy variable representing the original COOL measure (DCOOL1) takes the value of 1 after 29 September 2008.409 Similarly, the United States notes that the dummy variable representing the amended COOL measure takes the value of 1 after 23 November 2013, while the amended COOL measure actually came "into effect" on 23 May 2013.410
5.159.
Canada states that it used data for the initial impact of the original COOL measure as of the end of September 2008 since "the industry in the United States and Canada understood that the COOL measure would be expected to be in force."411 The model is designed to show the impact of the COOL measure based on incentives to industry participants, and "[b]ased on comments from USDA and members of Congress, industry participants understood that they would be expected to comply with the original COOL measure … as of the end of September 2008."412
5.160.
Regarding the amended COOL measure, Canada recalls that labelling was only enforced after 23 November 2013, six months after the announcement of the Final Rule.413 A "period of education and outreach … lasted for six months" from May to November 2013, delaying enforcement and resulting in "no incentive for retailers to use a complex and costly new labelling regime until after 23 November 2013."414

5.2.3.3.2 Analysis by the Arbitrator

5.161.
We note that the aim of Canada's and Mexico's econometric estimations is to observe the impact of the COOL measure on the export market for livestock. In other words, the methodology is intended to capture actual effects. These effects may or may not coincide with the formal entry into force of the measure.
5.162.
In respect of the specification for the original COOL measure, we recall the panel's finding in the original proceedings in these disputes that "the COOL measure started to develop its effect in 2008, shortly after the United States' economic recession started in December 2007."415 Furthermore, the United States has not rebutted Canada's assertion that industry participants had already begun to anticipate the COOL measure as early as September 2008, and started acting accordingly.416 Recalling the rules on burden of proof, the notion that econometrics is intended to capture actual market effects, and the lack of support provided by the United States for its assertion, we are satisfied with Canada's definition of the original COOL dummy.
5.163.
As for the amended COOL measure, the United States does not refute Canada's assertion of a period of "education and outreach" for the six months from May 2013 to November 2013, nor Canada's assertion regarding industry incentives to use the new, more costly labelling regime. We find Canada's explanation compelling and reject the United States' contention that the dummy variable representing the amended COOL measure was mis-specified.

5.3 CONCLUSION ON ASSESSMENT OF PROPOSED LEVEL OF SUSPENSION

5.164.
In this section we examined whether the United States successfully established that Canada's and Mexico's proposed levels of suspension are not equivalent to the level of nullification or impairment. We found this to be the case for the following reasons:

a. Canada's and Mexico's losses from domestic price suppression are not included in the nullification or impairment measured under Article 22 of the DSU; and

b. In respect of export revenue losses

i. Canada's econometric estimation of price basis does not account for price impacts from differential transport costs,

ii. Canada's invoice-based estimation of feeder pig prices is not reliable,

iii. Canada's econometric estimation of quantity does not adequately control for relevant explanatory variables, and

iv. Mexico's quantity simulation is based on an elasticity figure insufficiently supported by evidence.

5.165.
We note, however, that among these flaws that we identified, not all are fatal to the methodology used, but can be addressed. This concerns in particular the inclusion of additional variables in a price estimation and the calculation of an elasticity figure in a quantity simulation. With these considerations in mind we turn to our own determination of the level of nullification or impairment.

6 THE ARBITRATOR'S OWN DETERMINATION OF THE LEVEL OF NULLIFICATION OR IMPAIRMENT

6.1 INTRODUCTION

6.1.
As noted above in section 4, our mandate requires us to make our own determination of the level of nullification or impairment if we find that we cannot accept Canada's and Mexico's determinations.417 In section 5 above, we found this to be the case and therefore now proceed to our own determination.
6.2.
We recall that previous arbitrators, in devising their own approaches, have either based their approach on elements of the methodologies initially proposed by the parties418, or have followed an altogether different approach.419 As explained in section 4.2 above, we decided to examine all aspects of the methodologies used by Canada and Mexico in order to identify any valid elements that we could use in our own determination, if necessary. In our summary to section 5 above, we have identified these elements.
6.3.
As noted in section 3.3 above, the United States disagreed with the methodologies of both Canada and Mexico and proposed its own, alternative methodology to estimate the counterfactual export revenue loss, namely a partial equilibrium model in the form of an equilibrium displacement model (EDM).420 We now turn to examining this methodology in order to see whether it is a possible alternative to working with elements from Canada's and Mexico's methodologies. As we observed above in section 4, any determination of nullification or impairment, because it is based on assumptions, is necessarily a "reasoned estimate" relying on "credible, factual, and verifiable information".421 This being the case, no methodology is perfect. The goal is to provide a reasoned estimate that is as accurate as possible. In our assessment of Canada's and Mexico's methodologies we have already discussed the strengths and weaknesses of these approaches as well as the reliability of the data used for their calculations. Our analysis of the EDM proposed by the United States will assess the strengths and weaknesses of that methodology in order to weigh them against those already identified for Canada's and Mexico's methodologies. On the basis of this comparative assessment, we will adopt the approach that we consider best suited in this case to providing our own reasoned estimate that is as accurate as possible.

6.2 DESCRIPTION OF THE EDM METHODOLOGY

6.4.
We briefly describe the main elements of the EDM methodology used by the United States.
6.5.
An EDM is a partial equilibrium model representing a system of demand and supply relationships of a specific market or various markets forming supply chains – here the United States' livestock and meat markets. An EDM simulates the changes in prices and quantities in all the modelled markets that arise when the system equilibrium is displaced because of an exogenous shock – in this case the removal of the COOL measure and its associated compliance costs.422
6.6.
The United States models the US market for livestock (cattle and hogs) and meat (beef and pork) by specifying a multi-animal and multi-sector EDM.423 In particular, five stages of the cattle/beef and hogs/pork production and marketing chain are modelled: (1) farm: cow-calf and farrowing (i.efeeder cattle/pigs); (2) slaughter: finishing (i.e. fed cattle/hogs); (3) wholesale: packing (wholesale-level beef/pork); (4) retail (retail-level beef/pork) and (5) consumers. For each step in the production and marketing chain, up to four types of equations are used.424
6.7.
In its model, the United States considers the different elasticities and the policy change (i.e. the removal of the compliance costs of the COOL measure) as exogenous parameters. Such an approach is standard in an EDM, which rests on the assumption that the elasticities of the endogenous supply and demand relationships (i.e. prices and quantities) are known. An EDM also assumes that compliance costs of the COOL measure are known ("cost wedge")425.
6.8.
As for the elasticity parameters, the United States borrows the value of most of these parameters from previous research and academic literature. For the elasticities for which there is no readily available information, the United States makes additional assumptions in order to use the values of other elasticities as proxies. For instance, since there is no available information on the supply elasticities for US imports of feeder or slaughter animals, the United States assumes that the supply elasticities for US imports of feeder or slaughter animals – for which there is no available information – take the same value as the supply elasticity for US imports of wholesale meat imports.426
6.9.
As for the compliance costs, the United States derives the compliance costs of the original and amended COOL measures at the different stages of the supply/marketing chain using the estimates provided in the 2009 and 2013 Regulatory Impact Analyses (RIA) conducted by the USDA.427 In particular, the United States makes the assumption that the compliance costs to provide country-of-origin information on Canadian and Mexican livestock are the same as the compliance costs to provide such information for US-origin livestock.428

6.2.1 Arguments of the parties

6.10.
Canada and Mexico both criticize the United States' approach for relying on simulation where actual observed data are available to measure the impact of the COOL measure.429 Mexico suggests that EDMs "are not a standard approach for use in an ex post analysis when data are available."430 Similarly, Canada points out that "in an analysis where data on actual results caused by policy changes are observed, it is feasible and preferable to conduct an assessment of actual outcomes", for instance through the use of econometrics.431
6.11.
Furthermore, Canada and Mexico highlight a number of assumptions relied upon in standard EDMs including the EDM used by the United States. Canada notes that broad general assumptions include: perfect competition; that all sources of animals are used in all markets by all plants and firms; no substitution in processing; constant returns to scale; a "key assumption that implies only tiny changes on livestock of all origins"; perfect market equilibrium before and after the COOL measure; full equilibrium with the COOL measure in place; and "full equilibrium and all adjustment completed under the 'but for COOL' counterfactual".432 According to Canada, these assumptions fail to capture reality, resulting in an unrealistic estimate of nullification or impairment.433
6.12.
Canada and Mexico also take issue with the United States' choice of elasticities, arguing that the elasticities used in the model are "inappropriate to measure the fullimpacts"of removing the COOL measure.434 Canada and Mexico both assert that the United States' reliance on short-run elasticities is misplaced, since the EDM "assumes that a new full equilibrium is established in all markets and no further adjustment is underway."435 Canada and Mexico also challenge the United States' choice of wholesale meat import elasticities as a proxy for feeder and slaughter livestock import supply elasticities.436 Mexico notes in particular that the United States lacks any "economic rationale" to use wholesale meat import elasticity as a proxy for feeder and slaughter animal import elasticities.437 Canada also notes that the United States relies on elasticities that were "ironically" determined econometrically, but using "inappropriate" methods and decades-old data.438