Chevron on 2

The illogic of the Supreme Court of Canada’s approach to deference to administrative interpretations of law

Readers with some salsa experience will probably know that, while most of the world dances it “on 1”, in New York it is danced “on 2”. The steps and moves are more or less the same, but the sequence is different. Another dance that can be varied in this way, as we learn from the Supreme Court’s recent decision in Canada (Minister of Citizenship and Immigration) v Vavilov, 2019 SCC 65, is the notorious Chevron two-step. As with salsa, one can prefer one style or the other. But, for what it’s worth, I find Vavilov’s “on 2” version of Chevron to be rather offbeat.


In Chevron USA Inc v Natural Resources Defense Council, Inc, 467 US 837 (1984), the US Supeme Court explained how courts were to review administrative decision-makers’ interpretations of what in Canada are sometimes called their “home statutes”:

When a court reviews an agency’s construction of the statute which it administers, it is confronted with two questions. First, always, is the question whether Congress has directly spoken to the precise question at issue. If the intent of Congress is clear, that is the end of the matter; for the court, as well as the agency, must give effect to the unambiguously expressed intent of Congress. If, however, the court determines Congress has not directly addressed the precise question at issue, the court does not simply impose its own construction on the statute, as would be necessary in the absence of an administrative interpretation. Rather, if the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency’s answer is based on a permissible construction of the statute. (842-43; footnotes omitted)

The first step, in other words, is to determine whether the statute is so vague or ambiguous as to require an exercise of interpretive discretion by the administrative decision-maker. The second step, taken if―and only if―the statute does call for such an exercise of discretion, is to review the administrative interpretation for reasonableness, and defer to it if it is not unreasonable.

There are some exceptions to this two-step analysis. For one thing, under United States v Mead Corp, 533 US 218 (2001), courts ask whether the administrative agency was meant to conclusively determine questions of law in the first place. This is sometimes known as “Chevron step zero”. For another, following FDA v Brown & Williamson Tobacco Corp, 529 US 120 (2000), certain questions are seen as too important for their determination to have been delegated to administrative agencies implicitly; nothing short of explicit Congressional command will trigger deference. But, at least where the administrative decision-maker is seen as authorized to make legal determinations, Chevron dictates ― for now anyway ― the normal approach.

Or, if you prefer seeing and hearing instead of reading, here’s how NYU students explained it a few years ago:


Now, compare this to the Vavilov framework. It begins with a fairly close equivalent to “Chevron step zero”. In cases where the legislature wanted the courts, and not administrative tribunals, to decide legal questions, whether by explicitly providing for correctness review or by creating an appeal from from the tribunal to a court, the courts must not defer. Nor will there be deference on (some) constitutional questions and “general questions of law that are ‘of central importance to the legal system as a whole'” [58, quoting Dunsmuir v. New Brunswick, 2008 SCC 9, [2008] 1 SCR 190 [62]]. This is somewhat analogous to the “important questions” exception in the United States, although Canadian “questions of central importance” may well be different from the American “important questions”. (I don’t think, for instance, that under Vavilov it is enough for a question to be “of deep economic and political significance [and] central to [a] statutory scheme”: King v Burwell (2015) (internal quotation omitted).)

But then, Chevron‘s two main steps are reversed. Subject to the legislative assignment and central questions exceptions applying, Vavilov says courts are to defer to administrative interpretations of law:

Where a legislature has created an administrative decision maker for the specific purpose of administering a statutory scheme, it must be presumed that the legislature also intended that decision maker to be able to fulfill its mandate and interpret the law as applicable to all issues that come before it. Where a legislature has not explicitly prescribed that a court is to have a role in reviewing the decisions of that decision maker, it can safely be assumed that the legislature intended the administrative decision maker to function with a minimum of judicial interference. [24]

This is, more or less, Chevron‘s step two. At this stage, no factor other than the existence of the administrative decision-maker, the absence of a legislative indication that courts must nevertheless be involved, and the non-centrality of the question at issue are relevant.

But then, Vavilov seems to suggest that, once it embarks on reasonableness review, the court needs to examine the statute at issue more closely ― to engage what co-blogger Mark Mancini has described as a “legal ‘hard look’ review”, including to determine whether there is actually the sort of ambiguity that, under Chevron, justifies deference to the administrative interpretation. Vavilov stresses that “while an administrative body may have considerable discretion in making a particular decision, that decision must ultimately comply ‘with the rationale and purview of the statutory scheme under which it is adopted'” [108, quoting Catalyst Paper Corp v North Cowichan (District), 2012 SCC 2, [2012] 1 SCR 5, [15]] and, further, “with any more specific constraints imposed by the governing legislative scheme”. [108] Crucially, Vavilov insists that

If a legislature wishes to precisely circumscribe an administrative decision maker’s power in some respect, it can do so by using precise and narrow language and delineating the power in detail, thereby tightly constraining the decision maker’s ability to interpret the provision. Conversely, where the legislature chooses to use broad, open-ended or highly qualitative language … it clearly contemplates that the decision maker is to have greater flexibility in interpreting the meaning of such language. … [C]ertain questions relating to the scope of a decision maker’s authority may support more than one interpretation, while other questions may support only one, depending upon the text by which the statutory grant of authority is made. [110]

This, by my lights, is Chevron‘s step one. In some cases, the Supreme Court says, the legislature leaves the administrative decision-maker with the latitude to choose among competing possible interpretations. But not always. To quote Chevron again, “[i]f the intent of [the legislature] is clear, that is the end of the matter; for the court, as well as the agency, must give effect to the unambiguously expressed intent of [the legislature]”.

I should note that this might not be the only way to read Vavilov. Paul Daly, for example, is quite skeptical of “intrusive reasonableness review” that would occur if courts take too seriously the admonition about there being, sometimes, only one interpretation of administrative decision-maker’s grant of authority. But, as Mark shows, this is certainly a plausible, and at least arguably the better reading of Vavilov. I may return to the debate between these readings in a future post. For now, I will assume that the one outlined above is at least a real possibility.


As already mentioned, this reversal of the “Chevron two-step” makes no sense to me. I find it odd to say that reviewing courts must start from the position that “respect for [the] institutional design choices made by the legislature” in setting up administrative tribunals “requires a reviewing court to adopt a posture of restraint on review”, [24] but then insist that respect for legislative choices also requires the courts to be vigilant in case these choices leave only one permissible interpretation. The view, endorsed in Dunsmuir, that deferential judicial review reflects the inherent vagueness of legal language, was empirically wrong (and indeed implausible, as I argued here), but coherent. The recognition in Vavilov that statutory language is sometimes precise and can have a definitive meaning is welcome, but it is logically incompatible with an insistence on deference and judicial restraint.

If the Vavilov court had wanted to limit deference to cases of genuine interpretive uncertainty, it ought to have followed Chevron in clearly asking courts, first, to identify such cases, and then, and only then, to defer. That, of course, runs the risk of deference being relatively rare ― a risk highlighted by Justice Scalia in a lecture on “Judicial Deference to Administrative Interpretations of Law“:

One who finds more often (as I do) that the meaning of a statute is apparent from its text and from its relationship with other laws, thereby finds less often that the triggering requirement for Chevron deference exists. It is thus relatively rare that Chevron will require me to accept an interpretation which, though reasonable, I would not personally adopt. (521; emphasis in the original)

Conversely, if the Vavilov court was serious about deference-across-the-board being required as a matter of respect for legislative choice, it should have doubled down on the earlier view that statutory language inherently fails to determine legal disputes. This, in my view, would have been madness, but there would have been method in’t.

The trouble is that, as I said in my original comment on Vavilov, the majority opinion is a fudge. Collectively, the seven judges who signed it probably could not agree on what it was that they wanted, other than a compromise, and so did not want anything in particular. And so we get a judgment that, in a space of three short sentences, requires judicial review to embody “the principle of judicial restraint” while being “robust”, [13] and insists on deference while stressing that there may well be only one reasonable opinion to defer to.


Different people, and different legal cultures, will find their own ways to dance to the same tune of judicial resignation before the administrative state. Perhaps we should regard their different solutions as mere curiosities, objects of wonder but not judgment. But I don’t find this new Canadian hit, Chevron on 2, especially elegant or exciting. Not that I am a devotee of the on 1 original; but its steps at least come in a logical sequence. The on 2 version demands, as it were, that judges step forward and backward at the same time, and, with all due respect to the Canadian judiciary, I am not sure that it ― or, anyone else, for that matter ― is quite capable of such intricate footwork. Toes will be crushed, and partners disappointed if not injured, before someone realizes that the music needs, at long last, to stop.

The Dark Art of Deference

Dubious assumptions of expertise on home statute interpretation

Mark Mancini, Clerk at the Federal Court of Canada

The views expressed in this post are the author’s own and do not reflect, in any way, the views of the Federal Court of Canada.

The 10th anniversary of Dunsmuir presents an opportunity to revisit perhaps its most controversial aspect: the seeds it planted for a presumption of deference on home statute interpretation. As Professor Daly notes, the presumption is a “black hole” which engulfs questions of statutory interpretation in administrative law: Paul Daly, “Unreasonable Interpretations of Law” in Judicial Deference to Administrative Tribunals in Canada at 235-36. To the Supreme Court, the presumption exists because  “expertise …inheres in a tribunal itself as an institution” (Edmonton East, at para 33).

In this contribution, I will engage with what I call the “dark art of deference” that Dunsmuir entrenched: the seamless transfer that courts make from expertise in policy to expertise in other matters, including statutory interpretation (Edmonton East, at para 83). The mere fact of membership in a shadowy guild of expertise does not a legal expert make.

Here, I aim to: (1) demonstrate that expertise writ large does not provide a sound justification for deference on questions of law, unless incorporated into the decision-maker’s enabling statute and (2) relatedly, argue that deference is not prescribed by extralegal justifications such as expertise, but only by statutory language, which determines the leeway a court should afford to a decision-maker. The presumption of deference, which could run against statutory language, should be abandoned.

My comments start from three propositions which are rooted in constitutional theory: (1) absent constitutional objection, legislation binds; (2) administrative decision-makers enabled by statute can only go so far as their home statute allows (3) it is a court’s job, on any standard of review, to enforce those boundaries; in American terminology, to “say what the law is” (Marbury v Madison; Edmonton East, at para 21).

***

As early as the seminal case of CUPE v NB Liquor, “expertise” has been used to justify deference on questions of home statute interpretation. Yet beyond this general proposition, Canadian administrative law history “reveals…that little work has been done to pinpoint exactly what the concept of expertise means…”: Laverne Jacobs and Thomas Kuttner: “The Expert Tribunal.” Perhaps as a result, the problems associated with expertise as a basis for deference have not been extensively explored.

Most basically, expertise may be a valid practical justification for deference, but it is hardly doctrinally valid. To treat a decision-maker’s decision as binding on the basis of expertise abdicates the role of courts in enforcing statutory boundaries. While the court still nominally retains this power under the presumption of reasonableness, the presumption is virtually irrebuttable.  As Justice Scalia once wrote:

If I had been sitting on the Supreme Court when Learned Hand was still alive, it would similarly have been, as a practical matter, desirable for me to accept his views in all of cases under review, on the basis that he is a lot wiser than I, and more likely to get it right. But that would hardly have been theoretically valid. Even if Hand would have been de facto superior, I would have been ex officio so. So also with judicial acceptance of the agencies’ views…

This criticism, compelling as it is, has not been so for Canadian courts, with Edmonton East as proof positive. Courts on judicial review do view expertise as a valid doctrinal reason for deference, and are willing to put aside their own interpretation of a statute in favour of a decision-maker’s. This could be defensible, at least practically, if courts deferred on the basis of expertise which existed empirically. But in the post-Dunsmuir world, courts do not investigate expertise empirically before determining whether the presumption of deference applies. The presumption, though justified by expertise, can exist despite it. Yet as Professor (now Premier) MacLachlan says, if expertise is to support a presumption of deference, interrogation of a decision-maker’s expertise cannot be avoided indefinitely.

In the “pragmatic and functional era” which predated Dunsmuir, the courts did not avoid the question of expertise in determining the standard of review. Even though privative clauses were given less emphasis as clear statutory signals (see Pezim and the discussion in Khosa, at para 87), there was still a focus on determining relative expertise. For example, in Pasiechnyk, the Court was faced with a question of home statute interpretation. The Court considered the composition, tenure, and powers of the Saskatchewan Workers’ Compensation Board before holding that the decision-maker was expert on the matter before it. In Pushpanathan, at para 33, the Court noted that “expertise must be understood as a relative, not an absolute concept” and reasoned that courts must conduct an analysis in each case to compare its expertise to the decision-maker’s on the matter before it.

Contrast Edmonton East, where there was an assumption of the decision-maker’s expertise, but no actual analysis. Professor Sirota calls this “post-truth jurisprudence” and justifiably so—the dissent noted that the decision-maker did not have expertise in matters of statutory interpretation, and noted that expertise acted as an unjustified “catch-all” for deference (Edmonton East, at paras 82-83).

The Edmonton East dissent points out the problem with an automatic presumption of deference.  The role of courts on judicial review, to have the last word on the statutory boundaries of administrative decision-makers, is limited to reviewing whether the presumption is rebutted. Unfortunately, Dunsmuir, let alone Edmonton East, gave little guidance on when the fortified presumption of home statute interpretation could be rebutted: see Justice Cromwell’s reasons in Alberta Teachers. If legislative intent is the “polar star” of judicial review (CUPE v Ontario (Minister of Labour), at para 149), the search for that intent is attenuated by a game of categorical whack-a-mole with the correctness categories set out in Dunsmuir.

A presumption could be a useful legal device if it were true that most administrative decision-makers were expert in matters of statutory interpretation. Perhaps, it could be argued, decision-makers develop “field sensitivity” in legal decision-making even if they are not expert at the outset.  As Professor Green notes, a presumption along these lines could save judicial review courts the transaction costs associated with conducting an exhaustive review of the expertise of decision-makers in every case.

But such a presumption is only useful if (1) the facts supporting its existence are generally, empirically true; (2) courts themselves actually conduct the analysis to determine whether the presumption is rooted in fact, as in the Pushpanathan analysis (assuming that something like “field sensitivity” could be measured empirically) and (3) there is clear guidance on how to rebut the presumption. None of these exist with the current presumption of reasonableness.

***

What is the alternative if expertise writ large is simply a bad proxy for deference in the post Dunsmuir world? The answer is that expertise could be a justification for deference, but it must be based in a statute which recognizes it with respect to the precise legal issue before the court. Professor Martin Olszynski argues that the court should adopt a presumption of correctness for home statute interpretation, based on the principle of legislative supremacy. I agree that legislative supremacy is a better justification for deference than an unsupported concept of expertise, but the simple fact that a decision-maker has been delegated statutory power is not enough to justify deference, and it is not enough to demonstrate that a decision-maker is expert. What matters is what the legislature says about the merits of the matter before the court, including whether the enabling statute codifies any indication of expertise on that matter.

The Supreme Court does not accept this position. In Khosa, at para 25 the Court noted that the mere fact of  delegation to a decision-maker justifies deference on the basis of generalized expertise. But this is imprecise. Simply because the legislature delegates power to a decision-maker says nothing about how intensely a court should review that exercise of power and it says nothing about the expertise of the delegated decision-maker. Any logical leap from the mere fact of delegation to deference is a legal fiction. As Professor Vermeule notes in his splendid book Law’s Abnegation: “The delegation fiction is modern administrative law’s equivalent of the fiction that the Queen-in-Parliament still rules England—although she is bound always to act on the advice of her ‘ministers’ ” (30).

While the legal fact of delegation is irrelevant to deference, the terms of the delegation itself are relevant. This is because a legislature may speak in capacious terms, setting out requirements for expertise in a decision-maker and justifying a wide scope of options for a decision-maker. It may also speak in narrow terms, implying the opposite. In other words, the legislature can mandate deference through language, and the court on judicial review can recognize this deference, and apply that range to the decision taken by the decision-maker. This is nothing more or less than statutory interpretation by the judicial review court; and there is no need for a complex standard of review analysis to determine the intensity of review. The Supreme Court in McLean recognized as much: sometimes, statutory language will only permit one reasonable outcome. But in other situations, where the court finds that there are multiple permissible outcomes available, a decision within the range will be legal.

This approach is consistent with the role of courts on judicial review to have a final say on the law. It is also consistent with the notion of deference “itself a principle of modern statutory interpretation” (McLean, at para 40), because a court will not interfere with an administrative decision unless the principles of statutory interpretation honestly say so. This is demonstrated on the facts of McLean. The Court upheld the interpretation of the British Columbia Securities Commission’s home statute because it fell within the range of alternatives which the statutory language could bear. On the interpretation undertaken by the Court of the text, context, and purpose of the statute, a number of reasonable options were available to the Commission, and it chose one. This meant its decision was reasonable.

The McLean approach moves us away from imputed expertise in the post-Dunsmuir world to a real analysis of the intent of the legislature. Under this analysis, there is no need for an abstracted notion of expertise justifying a lawerly presumption of deference; nor is there a need for the labels of “reasonableness” or “correctness.” There is only a need for a court to defer where there is some indication in language to do so, including a statutory indication of expertise.

This is similar to the approach to judicial review of agency determinations of law in the United States under Chevron. Under Chevron, unlike Canada’s current standard of review framework, there is no specific analysis to determine the standard of review. For all its warts, Chevron at least avoids the protracted standard of review analysis, with its attendant standards, presumptions, categories, and factors which are confusing for the most learned of administrative lawyers. Its focus is properly on the enabling statute.

In fact, Dunsmuir and its progeny complicate the fact that judicial review on questions of law is fundamentally about statutory interpretation, nothing more or less. While there may be some trepidation associated with using the principles of statutory interpretation to discern whether deference is due, this is unavoidable. If administrative decision-makers are set up by statute, we cannot avoid an analysis of the decision-maker’s statutory powers, and the principles of statutory interpretation are the best we have. The key point is that the principles should not be used to determine if the statute is ambiguous or not, à la Chevron—they should be used to determine whether the statute’s language is sufficiently “open textured” to justify deference.

To come full circle, how does this approach square with the problem of expertise? In Edmonton East, the Court listed a number of justifications for a presumption of deference, including “access to justice” and expertise. These justifications exist outside of the statutory context, in the abstract. They are not necessarily tethered to the scope of the delegation afforded to the decision-maker. However, if statutory language is the focus, expertise can be a supporting justification for deference if represented in statutory language. The problem with the current presumption of deference is that it is based on expertise which may or may not be represented in statute.

Dunsmuir, 10 years on, has instantiated the dark arts of deference into the fabric of the Canadian law of judicial review in a manner untethered to statute. This is not a welcome development. Expertise should not be a “free standing” justification for deference (Khosa, at para 84). Perhaps it is time to pull back the curtain on the dark arts and drill down to the root of the law of judicial review: statutory interpretation.