World Trade Review

http://journals.cambridge.org/WTR

Additional services for World Trade Review:

Email alerts: Click here

Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here

The meaning of ‘necessary’ in GATT Article XX and GATS

Article XIV: the myth of cost–benet balancing

DONALD H. REGAN

World Trade Review / Volume 6 / Issue 03 / November 2007, pp 347 - 369

DOI: 10.1017/S1474745607003424, Published online: 31 October 2007

Link to this article: http://journals.cambridge.org/abstract_S1474745607003424

How to cite this article:

DONALD H. REGAN (2007). The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV:
the myth of cost–benet balancing. World Trade Review, 6, pp 347-369 doi:10.1017/
S1474745607003424

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/WTR, IP address: 129.49.23.145 on 14 Mar 2015

 World Trade Review (2007), 6 : 3, 347–369 Printed in the United Kingdom
f Donald H. Regan doi:10.1017/S1474745607003424

The meaning of ‘ necessary’ in GATT

Article XX and GATS Article XIV : the
myth of cost–benefit balancing
D O N A L D H. R E G A N *
William W. Bishop, Jr. Collegiate Professor of Law, University of Michigan

Abstract : Conventional wisdom tells us that in Korea–Beef, the Appellate Body

interpreted the word ‘necessary ’ in GATT Article XX to require a cost–benefit
balancing test. The Appellate Body is supposed to have applied this test also in
EC–Asbestos, US–Gambling (involving GATS Article XIV), and Dominican
Republic–Cigarettes. In this article I demonstrate, by detailed analysis of the
opinions, that the Appellate Body has never engaged in such balancing. They
have stated the balancing test, but in every case they have also stated the principle
that Members get to choose their own level of protection, which is logically
inconsistent with judicial review by cost–benefit balancing. And they have
decided every case by reference to the ‘ own level of protection ’ principle. The
Appellate Body is right not to balance. Balancing is not authorized by the treaty
texts, and it is not needed to prevent inefficient harm to foreign interests.

1. The argument in a nutshell

In Korea–Beef the Appellate Body interpreted for the first time the word ‘ necess-
ary ’ in GATT Article XX, specifically in paragraph XX(d).1 It is generally believed
that the Appellate Body established a cost–benefit balancing test : in order to decide
whether a measure is necessary to achieve some specified goal, we ‘ weigh and
balance ’ the benefits from the measure in the achievement of that goal against the
cost of the measure in reduced trade.2 I think this is a misunderstanding. The

* Email: [email protected].
Thanks to Rob Howse, Steve Charnovitz, Federico Ortino, Lorand Bartels, and an anonymous reviewer
for comments on previous drafts.
1 Korea – Measures Affecting Imports of Fresh, Chilled and Frozen Beef, WT/DS161 and 169/AB/R
(adopted 10 January 2001).
2 E.g., Marceau and Trachtman, ‘The Technical Barriers to Trade Agreement, the Sanitary and
Phytosanitary Measures Agreement, and the General Agreement on Tariffs and Trade: A Map of the
World Trade Organization Law of Domestic Regulation of Goods’, 36 Journal of World Trade (5, 2002)
811, at 826–828, 851–853; WTO, World Trade Report 2005: Exploring the Links between Trade,
Standards and the WTO, at 135, 137–139.

347

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 348 D O N A L D H. R E G A N

Appellate Body did not establish such a test. But I acknowledge that they made it
easy for people to think they did.3
The source of the confusion is a logical contradiction at the heart of the
Korea–Beef opinion. The Appellate Body does state the balancing test as I just
described it.4 But elsewhere in the opinion, the Appellate Body also states the
principle that a Member pursuing some legitimate domestic goal is entitled to
choose for itself the level of achievement of that goal. That is, a Member gets to
choose its ‘ own level of protection ’.5 Unfortunately, these two pronouncements of
the Appellate Body – that measures must be subjected to a balancing test and that
Members get to choose their own level of protection – are logically contradictory.
A court that is actually applying the balancing test described above must stand
ready to say of some measure that : (a) it achieves a legitimate local goal, and (b)
there is no other less trade-restrictive way to achieve the same level of that goal,
but (c) the measure is nonetheless illegal because the local benefits do not justify
the trade costs. But if the Appellate Body actually says that about some measure,
then the Member will be prevented from having the level of protection it has
chosen in adopting that measure. In case this is not obvious : any alternative
measure that achieves the same level of protection will involve equal or greater
trade cost, by (b) ; and any measure that achieves that level of protection at
equal or greater cost will fail the balancing test, by (c). So, applying a standard
cost–benefit balancing test is inconsistent with allowing the Member to choose its
own level of protection.
In the Korea–Beef opinion as a whole, the ‘ own level of protection’ idea gets
considerably more play than the balancing approach (even though the ‘balancing ’
passages have got much more attention from commentators). And if we look
specifically at the part of the opinion where the case is actually being decided, we
see that the Appellate Body relies on the principle that a Member gets to choose its
own level of protection.6 So the Appellate Body cannot actually be committed to
the cost–benefit balancing test described above, whatever they say.
There is a further source of confusion. The concrete test the Appellate Body
applies to implement the principle that Members get to choose their own level of
protection is a ‘ less-restrictive alternative test ’, which is just what is suggested
most naturally by the word ‘necessary ’. But their sophisticated version of the less-
restrictive alternative test itself involves a limited sort of balancing. For a start, the
Appellate Body considers as eligible alternatives to the actual measure only other

3 Among those who were not deceived are Howse and Tuerk, ‘ The WTO Impact on Internal
Regulations – A Case Study of the Canada–EC Asbestos Dispute’, in G. de Búrca and J. Scott (eds.), The
EU and the WTO: Legal and Constitutional Issues (2001), 283, at 324–325; F. Ortino, ‘ From Non-
Discrimination to Reasonableness: A Paradigm Shift in International Economic Law? ’, Jean Monnet
Working Papers 01/2005.
4 ""162–164, discussed in detail in Section 3.
5 ""176, 178, discussed in detail in Section 3.
6 "180, discussed in detail in Section 3.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 349

possible measures that would secure the same level of protection. (This guarantees
the Member’s right to choose its own level.) But within this class of alternatives,
the Appellate Body holds that a Member may have to adopt some alternative
measure that is less trade-restrictive than the actual measure, even though the
alternative measure involves higher administrative or enforcement costs. To decide
whether the Member must adopt the less trade-restrictive measure, the Appellate
Body balances the savings in trade cost achieved by the alternative measure
against the higher administrative/enforcement cost it entails. If the alternative
measure achieves the same level of the ultimate goal at lesser trade cost and
without unreasonable additional administrative/enforcement cost, the alternative
must be preferred, and the actual measure is not ‘ necessary’. Conversely, the
actual measure is ‘ necessary ’ if every alternative measure that achieves the same
level of the ultimate goal at lesser trade cost entails unreasonable administrative/
enforcement cost. This mode of decision obviously involves balancing of
administrative/enforcement costs against trade costs. But it is not the standard
cost–benefit analysis the Appellate Body describes in the most famous passages of
Korea–Beef, because it does not involve balancing the achievement of the under-
lying goal of the measure against the trade costs. As noted, the only alternatives
that are eligible for consideration are those that achieve the same amount of the
underlying goal.
The difference between these two sorts of balancing is enormously important.
To my mind, the most objectionable feature of Korea–Beef and the line of cases
that follow it is the suggestion that the Appellate Body is entitled to judge the
relative importance of various (non-protectionist) goals the Members might
have – human health, animal or plant health, preservation of species, avoidance of
noise pollution, restriction of luxury consumption, preventing offense to religious
sensibilities, whatever. There is no textual warrant for such judgments of import-
ance by the Appellate Body, and it is a serious intrusion on the Members’
regulatory autonomy.7 Of course, judging the importance of the local goal in this
way is unavoidable if the Appellate Body is actually going to engage in standard
cost–benefit balancing. But if the only balancing they engage in does not involve
balancing trade costs against the achievement of the underlying goal, but only
against administrative/enforcement costs, then there is never any need for the

7 The reader might object that Article XX itself implicitly ranks possible goals, by mentioning some
and not others. There is some truth in this, although the significance of the positive list in Article XX is less
than many people think, if we interpret Article III correctly – that is, as violated only by origin-specific
regulations and by origin-neutral regulations with a protectionist purpose – see Regan, ‘Regulatory
Purpose and ‘‘Like Products’’in Article III:4 of the GATT (With Additional Remarks on Article III:2) ’, 36
Journal of World Trade (3, 2002) 443, and ‘Further Thoughts on the Role of Regulatory Purpose Under
Article III of the General Agreement on Tariffs and Trade – A Tribute to Bob Hudec’, 37 Journal of World
Trade (4, 2003) 737. But even if Article XX does imply some ranking, it still does not authorize the
Appellate Body to do further ranking on its own, which is what the Appellate Body here proposes to do.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 350 D O N A L D H. R E G A N

Appellate Body to judge the importance of the underlying goal. This is a great
advantage of the less-restrictive alternative approach.
There are three later cases in which the Appellate Body has interpreted the word
‘necessary ’ in GATT Article XX or GATS Article XIV : EC–Asbestos8 (involving
GATT XX(b)), US–Gambling9 (involving GATS XIV(a), which corresponds to
GATT XX(a)), and Dominican Republic–Cigarettes10 (involving GATT XX(d)
again, as in Korea–Beef ).11 Unfortunately, the opinions in these later cases merely
reproduce the confusion and contradiction of Korea–Beef. All three opinions
quote faithfully from Korea–Beef both the statement of a standard cost–benefit
balancing test and the statement that each Member gets to choose its own level of
protection, with no recognition of the inconsistency. Furthermore, all three
opinions do so in such a way as to give prominence to the standard cost–benefit
balancing test, reinforcing the general perception that that is ‘ the test ’. Finally, all
three follow the lead of Korea–Beef and decide the actual case before them by a
‘less-restrictive alternative ’ test that respects the principle that Members get to
choose their own level of protection. Hence none of the later opinions actually
involves standard cost–benefit balancing, and none depends on a weighing of the
underlying benefit.
There is more to praise in these four opinions (Beef, Asbestos, Gambling, and
Cigarettes) than to criticize. I think the Appellate Body is right not to engage in
cost–benefit balancing, and the specific results are all defensible, even though I
might not have reached the same result as the Appellate Body in DR–Cigarettes.
But the doctrinal state of affairs – involving repeated statements of contradictory
tests side-by-side – is unfortunate and dangerous. It is unfortunate because we
ought to expect logical consistency from the Appellate Body. It is dangerous
because as long as the Appellate Body continues to state a standard cost–benefit
balancing test in part of each opinion on ‘ necessity ’, and as long as trade lawyers
continue to believe that such balancing is the applicable test, there is always the
possibility that the Appellate Body will be taken in by its own misdescription
of what it has been doing and will start actually trying to balance the domestic
benefits of a measure against the cost in reduced trade. That would be at odds with
the whole spirit of the WTO agreements, which leave the evaluation of domestic
benefits to domestic regulators.12

8 European Communities – Measures Affecting Asbestos and Asbestos-Containing Products, WT/

DS135/AB/R (adopted 5 April 2001).
9 United States – Measures Affecting the Cross-Border Supply of Gambling and Betting Services,
WT/DS285/AB/R (adopted 20 April 2005).
10 Dominican Republic – Measures Affecting the Importation and Internal Sale of Cigarettes, WT/
DS302/AB/R (adopted 19 May 2005).
11 There is passing mention of the ‘necessity’ issue under Article XX, with a cite to Korea–Beef and a
mention of balancing, in Mexico – Tax Measures on Soft Drinks and Other Beverages, WT/DS308/AB/R
(adopted 24 March 2006), "74. But there is no discussion of this issue, and it is perfectly clear that the
action in Mexico – Soft Drinks is elsewhere, so I shall ignore this case.
12 See Section 5.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 351

2. Logical preliminaries
Can the less-restrictive alternative test and cost–benefit balancing really
be inconsistent?
I have suggested that in Korea–Beef and its progeny, the principle that Members
get to choose their own level of protection is embodied in a sophisticated version of
the less-restrictive alternative test. And I have explained why this is inconsistent
with standard cost–benefit balancing. But the reader may still wonder how there
can be any logical inconsistency. In other contexts we think of the three versions or
stages of ‘proportionality’ analysis – the mere rationality test, the less-restrictive
alternative test, and cost–benefit balancing (sometimes known as ‘ strict pro-
portionality ’) – as forming a nested sequence of successively more stringent tests.
But if the less-restrictive alternative test and cost–benefit balancing are elements in
such a nested sequence, it must be possible to apply them both. Applying the less-
restrictive alternative test cannot exclude the possibility of cost–benefit balancing.
The seeming paradox goes away if we are careful to distinguish between positing
a test as a necessary condition for the legality of a measure and positing the
same test as a necessary-and-sufficient condition for legality. When we think of the
rationality test, the less-restrictive alternative test, and cost–benefit balancing
as forming a nested sequence, we are implicitly thinking of the first two tests as
stating merely necessary conditions for the legality of a measure. In order to be
legal, a measure must pass the minimum rationality test ; assuming it passes that,
then in order to be legal it must also pass the less-restrictive alternative test ; and
assuming it passes that, then in order to be legal it must also pass the cost–benefit
balancing test. Passing the balancing test is necessary-and-sufficient for legality
(unless there is some other unrelated test we have not mentioned), but the first two
tests are merely necessary conditions.13
In contrast, if we think of each of the tests as stating a candidate necessary-
and-sufficient condition for legality, then they do not form a nested sequence. They
are incompatible. In particular, if the less-restrictive alternative test states not
only a necessary condition for legality but also a sufficient condition, then the
cost–benefit balancing test cannot state a necessary condition, since it is easy to
imagine that some measure passes the less-restrictive alternative test but fails the
cost–benefit balancing test as the relevant tribunal would apply it.
But now the crucial point: By relying on the idea that Members get to choose
their own level of protection, the Appellate Body in Korea–Beef makes its version
of the less-restrictive alternative test necessary-and-sufficient for legality. If there is

13 The claim in the text that passing the balancing test is itself sufficient for legality presupposes that
the balancing test is applicable not only to the overall benefits and costs of the measure as compared to
inaction, but also to the (marginal) benefits and costs of the measure as compared to any alternative. For a
fuller treatment of the logic of this family of tests, see Appendix 2 to Regan, ‘Judicial Review of Member-
State Regulation of Trade Within a Federal or Quasi-Federal System: Protectionism and Balancing, Da
Capo’, 99 Michigan Law Review (8, 2001) 1853, at 1899–1902.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 352 D O N A L D H. R E G A N

no less trade-restrictive way to achieve the chosen level of protection without

unreasonable administrative/enforcement cost, the actual measure is legal, period.
There is no logical space left for cost–benefit balancing with the underlying goal.
So, it is not just the general statement of a less-restrictive alternative test that is
important in Korea–Beef ; it is the association of that test with the principle that
Members get to choose their own level of protection.

Does granting a margin of appreciation entail cost–benefit balancing?

I have said it is a great advantage of the less-restrictive alternative approach that it
does not require the dispute settlement system to evaluate the importance of the
regulator’s (non-protectionist) purpose. But we should consider a possible defense
of the Appellate Body’s claim to engage in such evaluation of purposes. Judgments
about whether some proposed ‘ less-restrictive alternative ’ actually achieves the
Member’s chosen level of protection can be controversial. So we are faced with the
question of what is the appropriate ‘margin of appreciation ’ that the regulator
should be allowed. It might be suggested that the appropriate margin of appreci-
ation varies with the importance of the regulatory goal. For example, we might say
the Appellate Body should be more deferential to the national regulator’s finding
that some alternative is inadequate when the regulation aims at an important goal,
so that the cost of mistakenly imposing an alternative that did not achieve the
chosen level of protection would be especially high. If this is right, the Appellate
Body obviously has to decide how important the regulatory goal is. Furthermore,
if this argument is accepted, it may seem to blur the distinction between the less-
restrictive alternative test and cost–benefit balancing, since the more important the
goal, the more likely it is that the regulation will prevail.
There are two principal points to make in response. First, even if the Appellate
Body behaves as suggested, that does not convert the less-restrictive alternative test
into cost–benefit balancing. Sometimes it is tolerably clear, without any attention
to the importance of the goal or the margin of appreciation, that there is no less-
restrictive alternative that achieves the chosen level of protection. In that case,
under a less-restrictive alternative test, the regulation is upheld ; under cost–benefit
balancing, it still faces another hurdle. Second, to the extent the Appellate Body
considers the importance of the purpose, the real question should not be about
the importance of the purpose in the abstract, or by the Appellate Body’s stan-
dards ; it should be about the importance of the purpose to the Member whose
regulation is in question. The proper measure of the harm done by an improper
invalidation is the damage to the Member’s purposes as it values them. Of course,
in assessing the importance of the purpose to the Member, the tribunal cannot help
but be influenced by views that are widely shared by other Members, or that are
endorsed by major multilateral treaties, or even by the tribunal’s own substantive
views. But the tribunal should do its best to remember that those are all merely
heuristics. The real question is about the importance of the purpose to the
Member, and if it is clear that the Member has an eccentric view, that view

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 353

should nonetheless be respected. The WTO imposes no general ranking of (non-

protectionist) goals.14
There is another way in which granting an extra margin of appreciation when
the regulatory purpose is important is not like traditional cost–benefit balancing.
Granting an extra margin of appreciation is a regulator-friendly move. Of course,
giving a greater margin of appreciation when the purpose is important logically
entails giving a relatively lesser margin of appreciation when the purpose is less
important ; but I have just pointed out that, even without any margin of appreci-
ation at all, if it is clear there is no less-restrictive alternative, the respondent
Member wins. So the margin of appreciation works in principle to the regulator’s
advantage. In contrast, traditional cost–benefit balancing, applied as a third-stage
test after the minimum rationality test and the less-restrictive alternative test, is
definitely not to the regulator’s advantage ; it is an additional hurdle. So the thrust
of the expanded margin of appreciation for ‘ important ’ purposes is entirely
opposite to the thrust of cost–benefit balancing. Incidentally, when we discuss the
Korea–Beef opinion in detail, we shall see that at crucial points, balancing is cast in
the role of a help to the regulator, not a hurdle. In context, there is another ex-
planation of that beside the present point about the margin of appreciation ; but, in
any event, this slant on the role of balancing supports my claim that the Appellate
Body is not imposing a standard cost–benefit balancing test.

3. The Korea–Beef opinion

I have argued that the Appellate Body in Korea–Beef posits, and claims to apply,
two contradictory tests. How did the Appellate Body get itself into such a mess ? I
do not have a complete answer, but we can at least watch it happen if we go
through the relevant portion of the Korea–Beef opinion step by step. We begin
with the famous "160, in which the Appellate Body observes that although ‘ the
word ‘‘ necessary’’ normally denotes something ‘‘that cannot be dispensed with or
done without, requisite, essential, needful ’’ ’, a ‘ standard law dictionary’ says that
in different contexts ‘ necessary ’ may mean anything from ‘ indispensable ’ to
merely ‘ convenient’.
The Appellate Body then continues in "161:
We believe that, as used in the context of Article XX(d), the reach of the word
‘necessary ’ is not limited to that which is ‘indispensable’ or ‘of absolute
necessity’ or ‘inevitable ’. Measures which are indispensable or of absolute
necessity or inevitable to secure compliance certainly fulfill the requirements of

14 With regard to whether Article XX itself imposes some ranking, see n. 7 above. Incidentally, there is
a threshold question, which arises before we even get to the ‘margin of appreciation ’ issue, whether the
claimed purpose of a measure is the Member’s true purpose. General views about the value of various
purposes, and thus the likelihood of Members’ pursuing them, are relevant here too, but with the same
caveat that they are relevant only as heuristics.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 354 D O N A L D H. R E G A N

Article XX(d). But other measures, too, may fall within the ambit of this ex-
ception.

In the last two sentences just quoted, the Appellate Body says (a) that a measure
which is indispensable is (‘certainly ’) ‘ necessary ’, and (b) that a measure may be
‘necessary ’ without being indispensable. I will come back to (b) in a moment. But
notice first that by asserting (a) the Appellate Body, before it even mentions bal-
ancing, has already excluded the possibility of a generally applicable cost–benefit
balancing test. There is no guarantee that a measure which is indispensable to
secure the regulator’s goal could not still fail a balancing test. But the Appellate
Body says that a measure which is indispensable ipso facto fulfills the requirements
of Article XX(d). So, there is no room for a balancing test when the measure is
indispensable.
In fact, asserting (a) does even more. In addition to logically excluding balancing
in some cases, it makes it implausible that we should ever want to engage in
standard cost–benefit balancing in any case at all, because the cases it excludes are
the only ones where such balancing could make any sense from a logical point of
view. Consider. We know that if the measure is indispensable, it is legal, period.
That leaves cases where the measure is not indispensable, that is, where there is
some alternative way to achieve the domestic goal to the same extent. Now there
are two subcases. One possibility is that the alternative measure that achieves the
same domestic benefit has the same or higher trade cost. Obviously we are not
going to require the Member to switch to the alternative measure in such a case,
and it does not require any balancing to tell us so. The other possibility is that the
alternative measure that achieves the same domestic benefit has a lower trade cost.
Now it is clear that we do want to require the use of the alternative measure, since
it secures the same benefit at lesser cost, and once again we do not need balancing
to tell us so. So there is no case in which we need to balance.
Actually, when I say it is clear we do want to require use of an alternative
measure that achieves the same benefit at lesser trade cost, I am ignoring admin-
istrative and enforcement costs, as we frequently do in abstract discussions. The
Appellate Body rightly recognizes later in its opinion that we cannot completely
ignore administrative and enforcement costs. It implies that we should not require
the use of the alternative measure, even though it achieves the same domestic
benefit at lesser trade cost, if the administrative and enforcement costs of the
alternative are excessive (""173,179, relying on findings of the Panel that the costs
would not be excessive in this case). In fact, the Appellate Body has already an-
ticipated this point in (b) above ; this is why a measure may be ‘necessary ’ even
though it is not indispensable. At this point the comparison that is called for be-
tween the actual measure and the alternative measure requires us to balance the
extra administrative/enforcement cost of the alternative measure against the saving
in trade cost from the alternative measure, but this is still not standard cost–benefit
balancing. Specifically, it does not require any weighing of the domestic benefit

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 355

that is the underlying goal of the measure (consumer protection, health, or what-
ever). By hypothesis the two measures being compared achieve the same level of
that benefit, so we can ignore it in comparing the two. In sum, there is no case in
which we actually need to engage in the sort of balancing that involves weighing
the underlying domestic benefit of the measure.
As the preceding paragraphs indicate, both (a) and (b) are distinctly regulator-
friendly moves. The assertion of (a) excludes cost–benefit balancing when the
actual measure is the least trade-restrictive way to achieve the desired level of
protection ; this is a regulator-friendly move because balancing is a more stringent
test than a ‘ less-restrictive alternative ’ or ‘ indispensability ’ test. But as we have
seen, (b) weakens the test still further, by saying that some measures may count as
‘ necessary’ even though they are not the least trade-restrictive way to achieve the
desired level of protection, if the administrative/enforcement costs of any less
trade-restrictive alternative are too great. It is a pity that the Appellate Body does
not pause at this point to explain (as I just have) why some measure might count as
‘ necessary’ even without being indispensable. They do explain this later on, but if
they had done it here the whole opinion might have played out differently.
However, instead of developing the logic of the two sentences we have so far
focused on, they turn (still in "161) to the realm of metaphor. They opine that
the term ‘ necessary ’ in Article XX(d) refers to ‘ a range of degrees of necessity’
constituting a ‘ continuum’ with the meaning ‘ indispensable ’ at one end and the
meaning ‘making a contribution to’ at the other. They then say that a ‘necessary ’
measure is ‘ located significantly closer to the pole of ‘‘indispensable ’’ than to the
opposite pole of simply ‘‘ making a contribution to ’’’.
The Appellate Body now begins "162 by reminding us of the special features of
Article XX(d). Article XX(d) is about measures designed to ensure compliance
with other laws and regulations, and the Appellate Body points out that ‘ XX(d) is
susceptible of application in respect of a wide variety of ‘‘ laws and regulations ’’ ’.
The Appellate Body famously continues:

It seems to us that a treaty interpreter assessing a measure claimed to be necessary

to secure compliance of [sic] a WTO-consistent law or regulation may, in appro-
priate cases, take into account the relative importance of the common interests or
values that the law or regulation to be enforced is intended to protect. The more
vital or important those common interests or values are, the easier it would be to
accept as ‘necessary ’ a measure designed as an enforcement instrument.

This is a surprising suggestion in a number of respects. First, it seems to constitute

a break in the line of reasoning. At the end of "161, the Appellate Body deployed
the metaphor of a ‘continuum ’ of ‘ degrees of necessity’ of a regulatory measure to
the achievement of a specified purpose. The problem was to locate ‘ necessary ’ on
that continuum. There is nothing in this metaphor that suggests an inquiry into the
value of the purpose itself, since the purpose is the same all along the continuum.
Furthermore, there is nothing in the text of Article XX(d) to suggest that different

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 356 D O N A L D H. R E G A N

regulatory purposes are accorded different values by Article XX(d). A fortiori,

there is nothing to suggest that it is appropriate for the Appellate Body to rank
Members’ regulatory purposes according to the Appellate Body’s intuitions about
their value.
Why does the Appellate Body, which has so often paraded its concern for the
text, make such a move ? I do not claim to know, but one might speculate that the
Appellate Body members, knowing they were going to emphasize each Member’s
right to choose its own level of protection later in the opinion, were worried that
if Members could choose their own level of protection with regard to just any
purpose at all, there would simply be too much regulatory freedom, too much
interference with trade (or from another perspective, too little Appellate Body
control). Leaving aside the question whether the Appellate Body should decide for
itself, independently of the text, how much regulatory freedom (or how much
Appellate Body control) there should be, this is surprising in a new way. As I have
explained, "161 of the opinion is notably regulator-friendly. But "162, on this
understanding, is notably regulator-unfriendly. Regulators are now not to be
trusted with the freedom they were conceded earlier. It is this ambivalence of the
Appellate Body about how much freedom regulators should have that generates
the contradiction I pointed out in the Introduction between the balancing test, first
intimated here in "162, and the idea that Members get to choose their own level of
protection, which was just stated, although not in those terms, in "161 and which
will be stated again in precisely those terms later in the opinion.
So, "162 seems, surprisingly, to introduce a standard cost–benefit balancing
test. If "162 stood alone, it might be taken to suggest merely that there should be a
greater margin of appreciation for regulations pursuing important purposes, as we
suggested in Section 2. This would be the charitable interpretation, since it avoids
introducing any logical inconsistency into the opinion. But the next two para-
graphs of the report make this ‘ non-balancing’ reading much less plausible:

163. There are other aspects of the enforcement measure to be considered in

evaluating the measure as ‘necessary ’. One is the extent to which the measure
contributes to the realization of the end pursued, the securing of compliance
with the law or regulation at issue. The greater the contribution, the more easily
a measure might be considered to be ‘ necessary ’. Another aspect is the extent
to which the compliance measure produces restrictive effects on international
commerce, [footnote omitted] that is, in respect of a measure inconsistent
with Article III.4, restrictive effects on imported goods. [Emphasis in original.]
A measure with a relatively slight impact upon imported products might
more easily be considered as ‘ necessary ’ than a measure with intense or broader
restrictive effects.

164. In sum, determination of whether a measure, which is not ‘ indispensable’,

may nevertheless be ‘ necessary’ within the contemplation of Article XX(d), in-
volves in every case a process of weighing and balancing a series of factors which

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 357

prominently include the contribution made by the compliance measure to the

enforcement of the law or regulation at issue, the importance of the common
interests or values protected by that law or regulation, and the accompanying
impact of the law or regulation on imports or exports.
By now the Appellate Body seems quite clearly to be stating a cost–benefit bal-
ancing test. To be sure we could still take the whole ‘ weighing and balancing ’
discussion of ""162–164 as just about the margin of appreciation in applying the
less-restrictive alternative test. But that is a strained reading – especially since the
Appellate Body never mentions the margin of appreciation – and it is not the way
this passage has been commonly understood. As I have explained in Section 2, if
this passage is just about the margin of appreciation in applying the less-restrictive
alternative test, that is very different from the cost–benefit balancing the passage is
usually thought to require.
Putting the margin-of-appreciation interpretation aside, we should note that
if the Appellate Body thought a balancing approach was necessary to account for
the significance of each of the three factors they name, they were mistaken. Both
of the considerations mentioned in "163 – the amount of the desired benefit
the measure achieves and the trade cost it imposes – are fully relevant to a less-
restrictive alternative test that involves no balancing. In order to know whether a
suggested ‘ alternative ’ to the actual measure achieves the same level of the benefit,
we have to know how much the actual measure achieves. And just as the Appellate
Body says in "163, the more benefit the actual measure achieves, the more likely it
will be found ‘necessary ’ (even without doing any cost–benefit balancing), because
the less likely it is that there is an alternative measure that is less trade-restrictive
and that achieves as much. Similarly, in order to know whether a suggested
‘ alternative ’ to the actual measure is less trade-restrictive, we need to know how
trade-restrictive the actual measure is. And once again, just as the Appellate Body
says in "163, the less trade-restrictive the actual measure is, the more likely it will
be found ‘ necessary ’ (even without doing any cost–benefit balancing), because the
less likely it is that there is an alternative measure that achieves the same benefit
and that restricts less. In fact, the only consideration in the Appellate Body’s list
that is relevant to a cost–benefit balancing test and not to a less-restrictive
alternative test is the value of the regulatory purpose, which as we have already
seen is a seriously suspect consideration.
Notice also, before we go on, that the contradiction that infects the opinion
as a whole is fully manifested in the famous sentence that constitutes "164. I
think most people remember that sentence as if it said : ‘ In sum, determination of
whether a measure is ‘‘ necessary’’, within the contemplation of Article XX(d),
involves in every case a process of weighing and balancing _ [of the factors
mentioned in ""162,163]. ’ But it does not say that. It says, ‘In sum, determination
of whether a measure, which is not ‘‘indispensable ’’, may nevertheless be
‘‘ necessary ’’ _ [involves balancing] ’ (emphasis added). Right here at the heart of
the supposed balancing test, we are reminded explicitly that the test does not apply

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 358 D O N A L D H. R E G A N

if the measure is actually indispensable. But as I explained already, this is the only
case where traditional cost–benefit balancing could actually make any contri-
bution to the analysis, as a logical matter. And there is a further revealing feature
of the syntax. The Appellate Body says, ‘ In sum, determination of whether a
measure, which is not ‘‘ indispensable ’’, may nevertheless be ‘‘ necessary’’ _
[involves balancing] ’ (emphasis added). The measure that is not indispens-
able is (by implication) presumptively not necessary, but it may nevertheless turn
out to be ‘ necessary’ ; it may be rescued, as it were, by a balancing inquiry. We
normally think of the balancing test as a hurdle the measure must get over, and
that is the right way to think about the standard cost–benefit balancing test, if we
apply it. But here balancing is portrayed implicitly as a life-preserver the measure
may be able to catch. This makes sense only if the balancing we are now talking
about is the sort described earlier as a kind of safety-valve on the less-restrictive
alternative test, which allows a measure to be found ‘necessary ’ even though there
is a less-trade-restrictive alternative available that achieves the same benefit, if the
alternative measure involves enough extra administrative/enforcement costs to
outweigh the savings in trade cost. This is the very test stated later in the opinion.
And as I have pointed out previously, this test requires no evaluation of the im-
portance of the underlying benefit pursued by the measure.
In "165 the Appellate Body leaves cost–benefit balancing and goes back to the
less-restrictive alternative test, although it is not immediately clear that this is what
is happening. The Appellate Body begins by quoting from the GATT Panel report
in US – Section 33715 on the meaning of ‘ necessary ’ in Article XX(d) :
[A] contracting party cannot justify a measure inconsistent with another GATT
provision as ‘ necessary ’ in terms of Article XX(d) if an alternative measure which
it could reasonably be expected to employ and which is not inconsistent with
other GATT provisions is available to it. By the same token, in cases where a
measure consistent with other GATT provisions is not reasonably available, a
contracting party is bound to use, among the measures reasonably available to it,
that which entails the least degree of inconsistency with other GATT provisions.
This sounds like a less-restrictive alternative test, not cost–benefit balancing, ex-
cept that the Appellate Body immediately says in "166 that this Section 337
standard
encapsulates the general considerations we have adverted to above. In our view
the weighing and balancing process we have outlined is comprehended in the
determination of whether a WTO-consistent alternative measure which the
Member concerned could ‘reasonably be expected to employ’ is available, or
whether a less WTO-inconsistent measure is ‘reasonably available’.

15 United States – Section 337 of the Tariff Act of 1930, BISD 36th Supp. 345 (adopted 7 November
1989), "5.26.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 359

So what is it, less-restrictive alternative or balancing? There is room for doubt

because of an ambiguity in the notion of an ‘ alternative ’ measure. If an eligible
‘ alternative ’ measure must be one which achieves the same level of the
benefit sought by the actual regulation, then we have a less-restrictive alterna-
tive test. There is no place for cost–benefit balancing, and there is no need for
a ranking of regulatory purposes. Any balancing to determine whether an
alternative is ‘ reasonably available ’ must be a matter of weighing the reduction in
trade cost it achieves (as compared to the actual measure) against any additional
administrative/enforcement cost. In contrast, we might just mean by an ‘ alterna-
tive ’ measure one which achieves ‘ some considerable amount ’ of the benefit
sought by the actual measure, or one which achieves ‘ almost as much ’ of that
benefit as the actual measure. And we would then most naturally mean by an
alternative that is ‘reasonably available ’, or that the Member ‘ could reasonably
be expected to employ ’, a measure whose saving in trade cost outweighs (in the
opinion of the Appellate Body) the reduction in the achievement of the Member’s
objective. This is cost–benefit balancing, and this does require the valuing by the
Appellate Body of Members’ regulatory purposes. So, what kind of balancing we
are doing depends on what we mean by an ‘alternative ’.
The Appellate Body allows this crucial ambiguity to persist for the next few
paragraphs, in which it works through the Korea–Beef Panel’s treatment of the
facts. Happily, we do not need to analyze these paragraphs in detail. At the be-
ginning of "176 the ambiguity about what counts as an ‘ alternative ’ is finally
resolved : ‘ It is not open to doubt that Members of the WTO have the right to
determine for themselves the level of enforcement of their WTO-consistent laws
and regulations. ’ This means, obviously, that a measure that does not achieve the
same level of protection is not an eligible alternative, whatever its other virtues.
And the Appellate Body points out that this was the view of the Section 337 Panel
also. They quote that Panel as saying that the obligation to choose a reasonably
available alternative measure ‘ does not mean that a contracting party could be
asked to change its substantive patent law or its desired level of enforcement of
that law ’(emphasis added by the Appellate Body).16 In other words, Members get
to choose their own level of protection (provided, the Section 337 Panel adds, that
they enforce the same level of protection against harm from imported and like
domestically produced products).
The next three paragraphs, ""177–179, also do not require detailed analysis.
They deal with the issue of what Korea’s chosen level of protection actually is, and
they can be summed up in the eminently sensible proposition that Korea cannot
assert a ‘ desired level of protection’ that is higher than the level achieved by the
actual measure and then hold proposed alternatives to the standard Korea asserts
as opposed to the standard its actual measure achieves. This is not to deny that
Korea may genuinely want a higher level than they are able to achieve. Even so, it

16 Korea–Beef, "176, quoting US – Section 337, "5.26.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 360 D O N A L D H. R E G A N

is obvious that a proposed alternative should not be required to achieve more

benefit than the actual measure.17
The Appellate Body finally summarizes in "180 its rationale for holding against
Korea, and it makes clear again that the operative test – the test on which it
explicitly relies to justify the result in the case – is the sophisticated less-restrictive
alternative test :
We share the Panel’s conclusion. We are not persuaded that Korea could not
achieve its desired level of enforcement of the Unfair Competition Act with re-
spect to the origin of beef sold by retailers by using conventional WTO-consistent
measures, if Korea would devote more resources to its enforcement efforts on the
beef sector.
In other words, Korea loses because the Appellate Body is not persuaded that they
could not achieve the same benefit at lesser cost to trade, if they just spent some-
what more (but, implicitly, not an unreasonable extra amount in light of the trade
cost avoided) on enforcement. This rationale involves no balancing with the fraud-
prevention benefit because the amount of that benefit achieved does not vary be-
tween the measures (actual and alternative) being compared. It follows that this
rationale requires no determination by the Appellate Body of the importance of
fraud-prevention.
In this connection it is worth mentioning that, although the Appellate Body
raises the issue of the importance of the regulatory goal in ""162,164, it never
actually tells us anywhere in the opinion how important it thinks Korea’s goal of
fraud-prevention is. Many readers probably assume the Appellate Body does tell us
about the importance of Korea’s purpose implicitly : the Appellate Body says the
importance of the Member’s purpose matters ; Korea loses ; so we can infer that
Korea’s purpose was relatively unimportant. If the opinion as a whole established
that the Appellate Body was balancing, this argument would be suggestive
(although not conclusive) about how the Appellate Body valued fraud-prevention.
But in fact the Appellate Body ultimately states a rationale that does not require
evaluation of the Member’s purpose. In this context, their failure to follow
through on the issue of the importance of the regulatory purpose simply confirms,
to my mind, that it is not a significant issue after all.18

17 The Appellate Body wrestles unsatisfactorily with a similar sort of question in Australia – Measures
Affecting Importation of Salmon, WT/DS18/AB/R (adopted 6 November 1998), part V.D. I am not sure
why they do not see what seems clear after the briefest reflection: that when we are applying a less-
restrictive alternative test, the ‘level of protection’ we should require from proposed alternatives is (a) the
level achieved by the actual measure, or (b) the level the Member asserts as its desired level, whichever is
lower (that is, whichever is less protective).
18 We shall see below that after deciding four cases on ‘necessity’, the Appellate Body has yet to say of
any legitimate purpose that it is not important; and in US–Gambling they seem to imply that fraud-
prevention in particular is important. Actually, the Appellate Body does seem to have a somewhat
jaundiced view of Korea’s purpose in the present case; but the best explanation is not that they see fraud-
prevention as unimportant. Rather, they suspect that Korea’s actual purpose is not origin-neutral
fraud-prevention. Remember the observation from the Section 337 Panel that the regulating Member can

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 361

In sum, the Appellate Body does not engage in cost–benefit balancing in

Korea–Beef. To be sure, they state a cost–benefit balancing test in three paragraphs
of the opinion (""162–164). But we have seen that they also state a different test,
which is logically incompatible with cost–benefit balancing, both before those
balancing paragraphs (in "161), and again after them (in ""176,180), and indeed
in the middle of one of them ("164). And it is this other test, based on the idea that
each Member gets to choose its own level of protection, that they explicitly rely on
in "180 when they actually decide the case.

4. Rehashing Korea–Beef: Asbestos, Gambling, and Cigarettes

Let us look now at the three Appellate Body reports that have followed
Korea–Beef. Happily, we will not need to parse any of them in the same detail as
Korea–Beef. The main point, as I said in the Introduction, is that all three repro-
duce the basic schematism of Korea–Beef. All three restate the Korea–Beef bal-
ancing test ; all three also state that Members are entitled to choose their own level
of protection; none notices that there is a logical contradiction here ; and when it
comes to actually deciding the case, all three rely on the principle that Members get
to choose their own level of protection.

EC–Asbestos
Consider first Asbestos. The Appellate Body begins by saying in "168 that
Members are entitled to choose their own level of protection: ‘ [I]t is undisputed
that WTO Members have the right to determine the level of protection of health
that they consider appropriate in a given situation. ’ A few paragraphs later, in
"172, the Appellate Body refers to Korea–Beef and its ‘ weighing and balancing
process ’:
We indicated in Korea–Beef that one aspect of the ‘weighing and balancing
process _ comprehended in the determination of whether a WTO-consistent
alternative measure’ is reasonably available is the extent to which the alternative
measure ‘contributes to the realization of the end pursued ’. In addition, we ob-
served, in that case, that ‘ [t]he more vital or important [the] common interests or
values ’ pursued, the easier it would be to accept as ‘necessary ’ measures designed
to achieve those ends. In this case, the objective pursued by the measure is the
preservation of human life and health through the elimination, or reduction, of

choose its own level of protection, but it must seek the same level of protection against harm from foreign
products and like domestic products. The Korea–Beef Appellate Body was plainly troubled by the fact that
Korea seemed to care less about domestic dairy beef being passed off as Hanwoo beef than about foreign
beef being passed off ("168). This looks like having a higher level of protection when the harm is from a
foreign product. (Related to the idea that Korea seems to want a higher level of protection against harm
from foreign beef is the Appellate Body’s complaint that the dual retail system puts ‘ all, or the great bulk’
of the costs of fraud-prevention on the foreign beef, as opposed to spreading the costs between domestic
and foreign beef ("181).)

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 362 D O N A L D H. R E G A N

the well-known, and life-threatening, health risks posed by asbestos fibres. The
value pursued is both vital and important in the highest degree. The remaining
question, then, is whether there is an alternative measure that would achieve the
same end and that is less restrictive of trade than a prohibition. [Footnotes
omitted.]
In "174, the Appellate Body tells us that Canada’s proposed alternative to France’s
measure, ‘ controlled use ’, would not allow France to achieve ‘ its chosen level of
health protection ’. And in "175 it tells us that ‘ for these reasons’, France’s
measure is ‘ necessary’ under Article XX(b) because there is no reasonably avail-
able alternative to the actual measure.
The one thing the Appellate Body does not do is show any awareness of the
logical contradiction between a balancing approach and the idea that France gets
to choose its own level of protection. So we should ask whether they have hit on
some way to make the contradiction go away. One possibility is that the Appellate
Body is saying that a WTO Member gets to choose its own level protection, pro-
vided the goal of the measure is ‘ vital ’ or ‘ important ’. So long as the goal is of this
favored sort, the Appellate Body does not balance, but with less important goals it
might. In support of this, it might be noted that whenever the Appellate Body
speaks most explicitly in Asbestos of France’s (or a Member’s) right to choose its
own level of protection, it speaks of the right to choose the level of health pro-
tection (""168,174). On the other hand, they never say explicitly that the right to
choose the level of protection is limited to cases involving health or life or other
‘important ’ values. Indeed, their first statement of the principle, in "168, although
it mentions health, otherwise closely tracks language from Korea–Beef and
US – Section 337, in which there is no such qualification. Also, US–Gambling cites
Asbestos for this idea without limitation to health.19 It seems most natural to
suppose that the Asbestos Appellate Body refers to health simply because that is
the regulatory goal the case involves, and not because they are trying to sneak in a
limitation on the Member’s fundamental right to choose its own level of protection
sub silentio.
Considering the ‘ important value ’ idea in its own right, there is reason to think
the Appellate Body does not take it all that seriously. In four cases the Appellate
Body has yet to say that any specific regulatory goal is not ‘ vital ’ or ‘ important ’. In
Asbestos, human life and health are vital and important. That’s easy. But in
DR–Cigarettes the Appellate Body apparently endorses the Panel’s view that col-
lecting tax revenue on cigarettes is vitally important.20 As I have already remarked,
people may read Korea–Beef as implying that preventing consumer fraud is not
within the Appellate Body’s charmed circle of preferred purposes, but the report
does not actually say any such thing. And in US–Gambling the Appellate Body

19 US–Gambling, "308 n. 379.

20 DR–Cigarettes, "71. The Panel did suggest that this sort of purpose might be specially important
for developing countries, but we can hardly make such a distinction part of the legal test of ‘ necessity’.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 363

makes no objection to the Panel’s finding that all of the United States’ purposes,
which include the prevention of consumer fraud, are important.21 We might of
course say that with four purposes involved in Gambling, the Appellate Body
wasn’t paying careful attention to each. Or we might say that the consumer fraud
in Gambling, which would presumably involve actual loss of money, was worse
than the consumer fraud in Korea–Beef, which involved getting perfectly edible
foreign beef instead of a preferred luxury domestic version. But the more hairs of
this sort the tribunals split, the more suspect the whole process would become. In
sum, the Appellate Body has yet to say that any specific legitimate regulatory
purpose is less valuable than any other. So far so good.

US–Gambling
In US–Gambling we see once again all the inconsistent elements of Korea–Beef.
First we read in "305, ‘ In Korea–Various Measures on Beef, the Appellate Body
stated, in the context of Article XX(d) of the GATT 1994, that whether a measure
is ‘‘ necessary’’ should be determined through ‘‘ a process of weighing and bal-
ancing a series of factors’’. ’ In "306, ‘ The process begins with an assessment of
the ‘‘relative importance ’’ of the interests or values furthered by the challenged
measures. ’ And so on through the Korea–Beef ‘ factors ’. The first two sentences of
"308 remind us that even if a less-restrictive measure is available, the Members
may not be required to use it if its administrative/enforcement costs are excessive.
Then the last sentence of "308: ‘ Moreover, a ‘‘ reasonably available ’’ alternative
measure must be a measure that would preserve for the responding Member its
right to achieve its desired level of protection with respect to the objective pursued
under paragraph (a) of Article XIV [of GATS] ’ (emphasis added). By this point,
the Appellate Body has transformed the cost–benefit balancing test into its logical
contrary – all unaware of the contradiction.
There is no single sentence in the Gambling report that encapsulates the decisive
rationale quite as neatly as in Korea–Beef or Asbestos, but the closest thing to such
a sentence is in "317 : ‘ In our view, the Panel’s ‘‘necessity’’ analysis was flawed
because it did not focus on an alternative measure that was reasonably available to
the United States to achieve the stated objectives regarding the protection of public
morals or the maintenance of public order’ (emphasis added). There are ten more
paragraphs in the ‘necessity ’ analysis, devoted to explaining why consultation
with Antigua was not, as the Panel thought, a relevant alternative, and why the
Appellate Body could then complete the analysis and find for the United States,
since no other alternative had even been suggested. These last ten paragraphs
talk about ‘ alternatives’ or ‘ reasonably available alternatives ’ without explicitly
addressing the ambiguity mentioned earlier about whether an ‘alternative ’ must
achieve the Member’s desired level of protection. But there is no reason to doubt

21 US–Gambling, "323.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 364 D O N A L D H. R E G A N

that this requirement, stated explicitly in "317, is continued implicitly throughout

the remaining discussion.
There is one interesting wrinkle in the Gambling report’s treatment of necessity.
Gambling is the only report that wrestles with the question of what counts as
a prima facie case of ‘ necessity’. ‘Necessity’ is of course part of a defense, so by a
standard principle, the burden is on the respondent Member to present at least
a prima facie case of ‘ necessity’. But if ‘ necessity’ depends on the non-existence of
a less-restrictive alternative, then a requirement to prove necessity is a requirement
to prove a negative, which another standard principle says we should not impose.
The Gambling Appellate Body is completely clear that the respondent Member
does not have to canvass and show the inadequacy of every imaginable alternative
measure (""309,320); at some point it is up to the complaining Member to suggest
‘reasonably available alternatives ’ (""320,326). But what must the respondent
Member do first? Something, surely, if there is to be any initial burden on the
respondent Member at all.
The Gambling report can be read as making the interesting suggestion that the
Korea–Beef ‘ balancing’ approach constitutes the respondent Member’s prima
facie case (""323–325). Once the prima facie case is made in this way, then we
move on to a less-restrictive alternative analysis that respects the respondent
Member’s right to choose its own level of protection and that requires the chal-
lenging country to suggest alternative measures. Now, this cannot be quite right. If
the prima facie case involves full-fledged cost–benefit balancing, we may cut off the
Member’s right to choose its own level of protection then and there, which by
hypothesis we do not mean to do. But there is a way to make some considerable
sense of the Appellate Body’s discussion here. Remember that two of the three
Korea–Beef factors – the degree to which the challenged measure advances the
regulatory goal and the degree of trade-restrictiveness – are fully relevant to a less-
restrictive alternative analysis.22 And if we ask what might constitute a prima facie
case of necessity (which is to say, of the absence of an eligible alternative), without
requiring the respondent Member to canvass all possible alternatives, it may be
that there is no better way to describe this than to say the Member must show a
high enough level of goal-achievement from the measure, and a low enough level
of trade-restriction, so that it seems plausible (pending the suggestion of alternative
measures by the complaining Member) that there is no alternative measure that
can achieve that much of the goal with less trade-restriction and at reasonable
administrative/enforcement cost. This might seem like a sort of proto cost–benefit
balancing (although it is not the real thing, since the desired level of protection is
respected).23

22 See Section 3 above.

23 Notice incidentally that the Appellate Body’s approach, as I have elaborated it, does not involve
shifting the final burden of proof on the ‘ necessity’ exception to the complaining Member. Once the
respondent Member has made a prima facie case for the necessity of its measure, as described in the text, it
is up to the complaining Member to propose some alternative; but once the complaining Member

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 365

DR–Cigarettes
Dominican Republic–Cigarettes adds nothing of interest with regard to the
necessity defense. Like the other cases, it restates the Korea–Beef balancing test,
citing also both Asbestos and Gambling (""66,68,69–70). It also states the in-
compatible principle that Members get to choose their own level of protection,
quoting from Gambling a sentence we have already noted : ‘ Moreover, a
‘‘ reasonably available ’’ alternative measure must be a measure that would pre-
serve for the responding Member its right to achieve its desired level of protection
with respect to the objective pursued’ ("70, quoting Gambling "308). The
Appellate Body then explains its ultimate resolution of the case in terms of the
latter principle. In "72 the Appellate Body summarizes thus the Panel’s reasoning
(which it then endorses in "73) :
[T]he alternative of providing secure tax stamps to foreign exporters, so that
those tax stamps could be affixed on cigarette packets in the course of their own
production process, prior to importation, would be equivalent to the tax stamp
requirement in terms of allowing the Dominican Republic to secure the high level
of enforcement it pursues with regard to tax collection and the prevention of
cigarette smuggling. [Emphasis added.]

Taking stock
The lesson of all four cases – Beef, Asbestos, Gambling, and Cigarettes – is that,
despite what the Appellate Body says in some parts of these opinions, it is not
engaged in cost–benefit balancing. Beef and Cigarettes may seem inconclusive on
their own ; in both cases the challenged measure fails the less-restrictive alternative
test, so we do not know for certain that the Appellate Body would not have gone
on to cost–benefit balancing if the measure had passed the less-restrictive alterna-
tive test. But in both Asbestos and Gambling, the measure passes the less-restric-
tive alternative test, and that is the end of the matter. Someone might claim that in
Asbestos the Appellate Body does balance, but there is no need to go through the
steps explicitly because the result is so obvious. In Gambling too, the hard-core
defender of balancing could insist that cost–benefit balancing is going on un-
consciously, or sub rosa. But despite the confusion the Appellate Body itself has
sown, the full opinions in all four cases strongly suggest that the determinative
principle is the principle that Members get to choose their own level of protection,
which excludes cost–benefit balancing. The claim that the Appellate Body is
‘ really ’ balancing, whatever it says, is the sort of claim it is impossible to disprove.
But if we are trying to account for a data set, we should accept the simplest ad-
equate explanation. The simplest adequate explanation for the results in these
cases is that the Appellate Body is deciding the cases by a less-restrictive alternative

proposes a minimally plausible alternative measure, the burden is still on the respondent Member to show
that the alternative does not achieve the desired level of protection, or is not less trade-restrictive, or
involves excessive administrative/enforcement costs.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 366 D O N A L D H. R E G A N

test, just as it says it is in the parts of the reports where the results are finally
announced and justified.

5. Why the Appellate Body is right not to balance: treaty text and
economic theory
I have argued that the Appellate Body does not engage in cost–benefit balancing. In
this section I shall explain briefly why they are right not to. First, they are right as a
matter of interpretation of the WTO texts. If we think about ordinary meaning in
context, the word ‘necessary ’ in GATT Article XX and GATS Article XIV suggests
a less-restrictive alternative approach much more naturally than it suggests a
cost–benefit balancing approach. Furthermore, we have seen that cost–benefit
balancing is incompatible with the principle that Members get to choose their own
level of protection. The right to set one’s own level of protection was recognized
under GATT before the WTO came into existence ;24 and the right is mentioned
specifically in the SPS and TBT Agreements.25 Starting from the other direction, if
we consider what sorts of limitations the WTO clearly imposes on Members’
regulatory autonomy, they are, roughly speaking : (a) the specific commitments
made in Members’ schedules under various agreements, (b) provisions in a number
of agreements designed to limit non-tariff protectionism (including subsidies),26
and (c) a ban on sanitary, phytosanitary, and technical measures that are more
trade-restrictive than necessary to achieve the regulating Member’s goals.27 None
of these limitations requires cost–benefit balancing, and there is nothing else in the
texts to suggest that the dispute system is authorized to subject measures to
cost–benefit balancing. Judicial review by cost–benefit balancing is not in the spirit
of the WTO.
Rejecting balancing is also right as a matter of economic theory. The proponent
of balancing may argue that, whatever the WTO texts say, we need balancing to
ensure global economic efficiency. Even a regulatory measure that is not protec-
tionist and not unnecessarily trade-restrictive may have greater costs abroad than
benefits at home (so the argument goes). Since the domestic regulator has no pol-
itical incentive to consider the costs abroad, the dispute settlement system must
balance the foreign costs against the domestic benefits to avoid inefficient results.

24 See the discussion of the Section 337 case in Section 3 above.

25 E.g., SPS 3.3, 4.1, 5.5, 5.6, TBT Preamble.
26 Notice that even subsidies to export industries, sometimes distinguished as ‘export promotion ’, can
be protectionism in the broad sense of regulation designed to improve the competitive position of some
group of domestic economic actors vis-à-vis their foreign competitors.
27 Of course there are numerous more specific disciplines in the SPS and TBT Agreements, but even
with regard to these, the ‘ least restrictive alternative’ principle is what they seem primarily designed to
implement, insofar as they are applied to non-discriminatory measures. And whatever the purpose of these
more specific disciplines, an inquiry into whether a measure is based on a risk assessment, or whether it is
based on an international standard, plainly does not require balancing.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 367

But this argument is fallacious. We do not actually need to balance the foreign
costs against the domestic benefits, not even to secure global efficiency. The reason
is that in the trade context (and given a general fact about countries’ policy-
making that I shall explain in the next paragraph), regulation by a Member that
maximizes that Member’s national social welfare is also globally efficient. I shall
summarize this in the motto that ‘ domestically rational regulation is globally ef-
ficient ’. This means that even if a tribunal reviewing some regulation is ultimately
concerned with global efficiency, all it needs to investigate is whether the regu-
lation is domestically rational. This does not trivialize the project of reviewing
regulation. Countries often adopt regulation that is domestically irrational – for
example, most protectionism, or well-motivated regulation that is unnecessarily
trade-restrictive (which is domestically irrational because it harms domestic con-
sumers unnecessarily, over and above harming foreign producers). There are
serious questions about how tribunals should go about identifying domestic
irrationality, and what sorts of domestic irrationality they can identify reliably
enough so that they should be charged to try. But still, if the issue is domestic
rationality, foreign costs are irrelevant. I cannot give an exhaustive demonstration
here of the claim that domestically rational regulation is globally efficient. I have
done so elsewhere.28 In a moment, I will offer an intuitive explanation. But first, let
me comment on some assumptions.
First, the claim that domestically rational regulation is globally efficient is lim-
ited to the trade context. It is obviously not true in general. If a Member allows, or
perhaps even requires, some local industry to dump noxious effluent into a river
that flows into a neighboring country, that may very well be domestically rational
but globally inefficient. But this is different from a trade law case, because the local
environmental law (or lack of it), creates a physical cross-border externality –
what an economist would call a ‘ non-pecuniary ’ externality. In contrast, trade
regulation creates only pecuniary externalities ; the harm done to foreign interests
is mediated through effects on market opportunities. This is a crucial difference.
Second, the claim that domestically rational regulation is globally efficient depends
on the assumption that countries do not choose trade-affecting regulation for the
purpose of improving their terms-of-trade. Notice I am not denying that trade-
affecting policies affect the terms of trade; I am only denying that they are chosen
on this ground. This is a controversial assumption in some quarters, but I have
argued for it in detail elsewhere.29 There are further assumptions we must make
before we can prove the claim in a mathematical model, but these two are the most
salient.30

28 Regan, ‘ What Are Trade Agreements For? – Two Conflicting Stories Told by Economists, With a
Lesson for Lawyers’, Journal of International Economic Law, 9(4), 951–988 (2006).
29 Ibid.
30 Other assumptions, aside from purely mathematical ones about convexity and the like, are (a)
competitive domestic markets, (b) constant returns to scale, and (c) domestically rational behavior by all

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 368 D O N A L D H. R E G A N

Even with these assumptions, why is it true that domestically rational regulation
is globally efficient ? Why do we not have to worry at some point about the foreign
costs ? Those costs are real even if they are only ‘‘pecuniary ’’. Over the years I have
experimented with a variety of metaphors and stories as intuitive explanations.
Here are two. First, suppose that Home imports widgets from Foreign. Although it
is individual consumers in Home who buy widgets, think of Home as a collective
consumer of widgets. The collective interest is just the aggregate of all private
interests in Home that are affected in any way by widget consumption in Home. If
there is some market failure in the domestic market – a consumption externality,
or consumer ignorance or irrationality, for example – then the demand curve for
imported widgets that results from the unregulated market behavior of Home’s
individual widget consumers will not embody the true interests of the collective
consumer Home. If Home now regulates rationally to correct the market failure, it
will alter the incentives of individual consumers so that their aggregated behavior
does reflect the true interests of the collective consumer. If we assume similarly that
Foreign puts in place any regulation that may be necessary to make the supply
curve for exports that results from the aggregated behavior of its producers reflect
the true overall interests of the collective producer Foreign, then the intersection
of Home’s demand curve for imports and Foreign’s supply curve for exports
will identify an efficient result, in the standard way. The one danger posed by
these ‘ collective ’ actors is that they will exploit their market power to manipulate
their terms of trade (that is, they will behave as monopsonist or monopolist
in the international market). But we have excluded that by hypothesis. So, if
each country regulates rationally for its domestic interests, the market produces
an efficient result. Domestically rational regulation is globally efficient.
The story we have just told emphasizes that separate self-interested regulations
by Home and by Foreign combine to produce an efficient result, via the mechanism
of the market. A different story (in which Home still imports widgets from
Foreign) emphasizes that Home’s widget consumers and Foreign’s widget pro-
ducers have a joint interest in being allowed to transact with each other on efficient
terms. Hence Home’s widget consumers can in a sense ‘ represent ’ Foreign’s widget
producers in Home’s political process, and vice-versa. Let us assume specifically
that consumption of widgets in Home creates a (domestic) consumption exter-
nality. (This is just one example to illustrate a general point.) To correct this
externality, Home imposes a Pigovian tax. But they overestimate the externality,
and they set the tax too high.31 Now, the too-high tax will create a deadweight
loss,32 and this loss will be divided between consumers in Home and producers

other countries. Again, I have explained elsewhere why these are appropriate assumptions for the dispute
settlement system to rely on, even though they are obviously idealizations. Ibid. at 982–983.
31 This is just a miscalculation by Home; by hypothesis, Home is not setting the tax above the efficient
level as a means of purposeful terms-of-trade manipulation.
32 This assumes there is efficient regulation in Foreign, so that Home’s too-high tax does not merely
compensate for, say, an export subsidy in Foreign.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145

 The meaning of ‘necessary’ in GATT Article XX and GATS Article XIV 369

in Foreign.33 That simple point is crucial : when the harm is market-mediated, any
loss is shared. The fact that the loss is shared means that if Home now revises its
policy so that it eliminates the deadweight loss on its own consumers, it will
necessarily eliminate the deadweight loss on Foreign’s producers as well. It’s not
that the loss of the consumers is the same as the loss of the producers; these losses
are different ; but they are linked. Because any deadweight loss is shared, reducing
the deadweight loss felt by Home’s consumers to zero means the deadweight loss
felt by Foreign’s producers must go to zero as well. In other words (and once
again), regulation that is domestically rational will also be globally efficient.
In eschewing cost–benefit balancing and allowing Members to choose their own
level of protection, the Appellate Body follows both treaty text and sound econ-
omic theory.

33 This assumes Home has market power as a buyer in the world market for widgets; if they do not,
their policy cannot hurt Foreign in any event.

http://journals.cambridge.org Downloaded: 14 Mar 2015 IP address: 129.49.23.145