Subject guide

NCA Administrative Law: How to Answer a Standard of Review Question

17 min read · NCA Edge

Administrative Law has a reputation among NCA candidates, and most of it is deserved. The subject is not conceptually harder than Constitutional or Criminal. What it punishes is a specific habit: learning the names of the cases without learning the order the analysis runs in. It costs most on the standard of review.

Almost every problem on an Administrative Law paper opens the same way. A tribunal or an official has decided something, somebody is unhappy about it, and your first job is to say how closely a court will look at that decision. Everything after that depends on your answer. The standard decides what you are even allowed to ask. Can you substitute your own reading of the statute? Or can you only test whether the decision-maker's reading holds together?

Here is the part that surprises people. On this question you can reach the right answer and still lose most of the marks. This guide walks the analysis in order, defines the terms it uses, shows the trap that catches strong students, and ends with a model paragraph and a checklist you can run under exam conditions.

First, two distinctions the rest of this depends on

If you trained outside Canada, these two may not map onto anything in your home system. Get them straight now, because everything below turns on them.

Judicial review is not an appeal. Judicial review is the courts' inherent supervisory power over administrative decision-makers. It exists whether or not any statute mentions it, and it cannot be taken away entirely. A statutory appeal is different: it exists only because a statute created it, and it reaches only as far as the statute says. Many Acts give a right of appeal on questions of law alone, or on questions of law and jurisdiction. Some give none at all. The two routes attract different standards, and telling them apart is the first move in the analysis.

The standard of review governs the merits, not the process. Everything in this guide is about how closely a court examines what the decision-maker decided. A separate question asks whether the decision was reached fairly — did the affected person get notice, a chance to respond, an unbiased decision-maker? That is procedural fairness, and Vavilov expressly leaves it outside this framework. On judicial review a court asks whether the procedure was fair in the circumstances, and it answers that without deferring to the decision-maker. Procedural fairness has its own body of law, and you should not run it through the standard-of-review analysis. There is one twist, and it comes later.

Step 1: start from the presumption

Since Canada (Minister of Citizenship and Immigration) v Vavilov, 2019 SCC 65, there is one starting point and it does not move. Whenever a court reviews the merits of an administrative decision, reasonableness is presumed.

Write that sentence down before you write anything else. It takes ten seconds and it frames the rest of the answer. The presumption rests on respect for the legislature's choice to give the decision to an administrative body rather than to a court. It applies whether or not the decision-maker was reading its own home statute — the Act that created it and gives it its powers.

Note what that removes. Expertise is no longer a factor you weigh when choosing the standard; it is already built into the presumption. Expertise still matters later, when you are actually conducting the reasonableness review, but it no longer decides which standard applies.

The presumption is displaced in two ways, and only two. Either the legislature said so, or the rule of law demands it. Every correctness category sits under one of those headings. Naming the heading as well as the category is a small move, and it shows you understand the structure rather than having memorised a list.

Step 2: work through the rebuttals

Route A: the legislature said so

The legislature prescribed the standard. This is rare, but it happens. If the statute states the standard a reviewing court must apply, that ends the question. Read the Act you are given. Examiners put the answer in the statute more often than you would expect.

The legislature provided a statutory right of appeal. This is the one that decides most exam problems, and the one most often reached by the wrong route. Where the statute gives a right of appeal from the decision to a court, the court applies appellate standards. It does not apply the Vavilov reasonableness framework at all. Those standards come from Housen v Nikolaisen, 2002 SCC 33, adopted for this purpose in Vavilov and applied in its companion case, Bell Canada v Canada (Attorney General), 2019 SCC 66.

To use them you have to sort each complaint into one of three kinds of question. Learn these definitions properly, because the whole route runs on them.

• A question of law asks what the law means. What does this statutory phrase cover? What is the legal test? The answer would govern other cases too.
• A question of fact asks what happened. Who said what, when, and what the evidence shows.
• A question of mixed fact and law asks whether the legal test is met on these particular facts. It is the two put together.

Sometimes a legal error is buried inside a mixed question. If the decision-maker used the wrong legal test, or misunderstood what a statutory phrase means, you can pull that error out and treat it on its own. That is an extricable question of law — extricable meaning, literally, that it can be separated out. If the complaint is only that the decision-maker applied the right test badly to these facts, there is nothing to extricate.

The standards then follow:

• Questions of law, including extricable questions of law — correctness. The court decides the question for itself and substitutes its own answer if it disagrees.
• Questions of fact, and questions of mixed fact and law with no extricable legal question — palpable and overriding error. This is a demanding threshold. "Palpable" means the error is obvious once pointed out, not something you need a re-trial to see. "Overriding" means it actually affected the result. A court that merely disagrees with the finding must leave it alone.

Two things about this route catch people out. First, an appeal right is often limited — "on a question of law", or "on a question of law or jurisdiction" — and it carries you only as far as its own terms. A complaint that falls outside the appeal right cannot be brought on appeal at all. Second, in Law Society of Saskatchewan v Abrametz, 2022 SCC 29, the Court held that on a statutory appeal these appellate standards apply even to procedural questions, in that case abuse of process through delay. That is the twist promised above: the usual separation of procedural fairness from the standard-of-review analysis does not survive a statutory appeal. Abrametz also shows the sorting in action — whether delay amounts to abuse of process is a question of law, reviewed for correctness, while the underlying findings of fact attract palpable and overriding error.

Route B: the rule of law

Three categories come from Vavilov itself. A fourth was added later, and it straddles both routes. All of them are narrow on purpose.

Constitutional questions. Vavilov points to three in particular. The division of powers between Parliament and the provinces. The relationship between the legislature and the other branches of the state. And the scope of Aboriginal and treaty rights under s 35 of the Constitution Act, 1982. The category is not closed; other constitutional matters fall in it too. A challenge to the constitutionality of the decision-maker's own enabling statute belongs here too.

Note the limit. This is not the same as an administrative decision that engages the Canadian Charter of Rights and Freedoms through the exercise of a discretion. Those remain governed by Doré v Barreau du Québec, 2012 SCC 12, which Vavilov expressly declined to reconsider. The distinction is worth holding onto. If the argument is that the statute is unconstitutional, you are in the correctness category. If the argument is that the decision struck the wrong balance with a Charter protection, you are in Doré.

General questions of law of central importance to the legal system as a whole. This is the category people reach for too quickly. It is not made out because a question is important to the parties, or difficult, or because it required reading a statute. It is made out where the answer has implications across the legal system and demands a single, consistent answer. The examples Vavilov itself gives are res judicata and abuse of process, the state's duty of religious neutrality, limits on solicitor-client privilege, and the scope of parliamentary privilege. If your fact pattern is a dispute about what one word in one Act means for one applicant, this is not your category.

Questions about the jurisdictional boundaries between two or more administrative bodies. Two tribunals, one dispute, and a question about which of them has the authority. Note the word "between". A single body arguing about the outer edge of its own powers is not here.

Questions of concurrent first-instance jurisdiction. "First instance" means the forum where a matter is decided for the first time, as opposed to on appeal or review. Sometimes a legal question can arrive first in either of two places — a tribunal or a court — because both have been given authority over it. That is what happened in Society of Composers, Authors and Music Publishers of Canada v Entertainment Software Association, 2022 SCC 30. The same Copyright Act provision could come before the Copyright Board, or before a court at first instance. Which one heard it first depended only on who started the proceeding. Correctness applies, because otherwise the same provision would mean one thing before the Board and another in court. The trigger is a genuinely shared first-instance forum. A question a court might simply find interesting is not enough.

SOCAN is also worth knowing as a structural point. Rowe J. grounded the new category on both headings — on legislative intent, because giving two bodies concurrent jurisdiction involves the courts by design, and on the rule of law. If you name this category, name both.

One thing to get right about all of Route B: correctness attaches to the question, not to the whole decision. SOCAN makes this explicit. The legal question in the category gets correctness, and the rest of the decision stays on reasonableness.

The trap: the right answer through the wrong door

The commonest way to lose marks here is to see something that feels important, conclude correctness, and then justify it with "a general question of law of central importance" — when the real trigger was sitting in the statute all along as a right of appeal.

Be clear about why that costs so much. It is not a labelling slip. The two routes are different tests with different consequences. Under the statutory appeal route you must sort every complaint into law, fact, or mixed, and attach a different standard to each. Palpable and overriding error is a demanding threshold, and skipping it means skipping the analysis the problem was built to produce. Under Route B you identify the particular legal question in the category and give that question correctness. Writing the second when the facts call for the first states a rule that does not govern, and then applies it.

Why this matters for marks: NCA answers are scored on whether you identify the issue, state the governing law accurately, and apply it to the facts. Your bottom line is not what earns the marks. So on this question the route is the answer. An examiner who reads the word "correctness" learns nothing about you. An examiner who reads why correctness applies learns everything. We go further into this in our guide on how NCA exams are marked.

The reverse error is harder to spot. Faced with a statutory appeal, some answers note it, cite Vavilov, and then review the whole decision for reasonableness anyway. That is not a matter of emphasis. On a statutory appeal, within the terms of the appeal right, Vavilov reasonableness is not the framework.

One more point converts a trap into marks. In Yatar v TD Insurance Meloche Monnex, 2024 SCC 8, the Court held that a limited right of appeal does not extinguish judicial review of everything else. Say the appeal right covers questions of law only. The remaining complaints can still be judicially reviewed, and a court's discretion to refuse to hear that review is narrow. The existence of the limited appeal is not itself a good reason to decline. A well-drafted problem can therefore put both frameworks in play at once, and the strongest answers say so.

Step 3: if it is reasonableness, what are you asking?

Answers that correctly land on reasonableness often then write a paragraph amounting to "this seems sensible enough to me." That is not the test, and it reads as an assertion.

Vavilov reasonableness is reasons-first. You begin with the decision-maker's actual reasons. You read them as a whole, with sensitivity to the administrative setting, and you ask whether the decision is justified, transparent and intelligible. In plain terms: do the reasons actually explain why, can you follow the path they took, and does that path make sense? What you do not do is work out the answer you would have reached and then measure the distance.

Mason v Canada (Citizenship and Immigration), 2023 SCC 21, is the sharpest recent application. The decision there was unreasonable because the decision-maker failed to grapple with central arguments and legal constraints put to it. The Court got there by reading the reasons, not by starting from its own preferred interpretation. Mason did go on to conclude that the constraints left only one reasonable interpretation open, and declined to send the matter back. But that was a remedial conclusion reached at the end. It is not a way of starting.

A decision is unreasonable if it has either of two kinds of fundamental flaw:

1. A failure of rationality internal to the reasoning. The chain of analysis does not hold together on its own terms.
2. It is untenable in light of the relevant factual and legal constraints that bear on it.

Those constraints are the part to learn, because naming the specific one your facts engage is where the application marks are. Vavilov lists them as illustrative rather than exhaustive, and says they are not a checklist and may vary in significance:

• the governing statutory scheme;
• other relevant statutory or common law;
• the principles of statutory interpretation;
• the evidence before the decision-maker;
• the submissions of the parties;
• the past practices and decisions of the administrative body;
• the potential impact of the decision on the individual affected.

That last one does real work where the stakes are high. The more severe the consequences for the person, the more responsive the reasons have to be.

One gap worth knowing. Sometimes there are no reasons, because none were required or none were given. Reasonableness still applies. The court then looks at the outcome itself against those same constraints, and asks whether it was a reasonable one.

Four things that are no longer the law

Marks leak here quietly, because none of these produces an obviously wrong sentence. They produce a confidently out-of-date one.

The Dunsmuir contextual factors. Dunsmuir v New Brunswick, 2008 SCC 9, still matters historically, but its four-factor balancing no longer decides the standard. Those factors were the purpose of the tribunal, the nature of the question, the decision-maker's expertise, and the presence of a privative clause. A privative clause is a provision by which the legislature tries to shield a decision from review by the courts. Working through those four factors now answers a question nobody asked.

"True questions of jurisdiction." This was a category of question about whether the decision-maker had the authority to enter on the matter at all, and it used to attract correctness. Vavilov abolished it. Arguing that a decision-maker "exceeded its jurisdiction" and that this triggers correctness is a reliable way to date your reading.

Correctness because it is "a question of law" or "outside the home statute." Neither is a category. A question of law decided by a tribunal, under its own statute or another one, is presumptively reviewed for reasonableness unless one of the routes above is engaged.

The old threshold for subordinate legislation. Regulations and by-laws used to be tested for validity under the Katz Group approach, which asked whether the measure was irrelevant, extraneous or completely unrelated to its statutory purpose. In Auer v Auer, 2024 SCC 36, and its companion TransAlta Generation Partnership v Alberta, 2024 SCC 37, the Court held that Vavilov reasonableness is the presumptive standard for reviewing whether subordinate legislation is authorised by its enabling statute. This is recent and very examinable.

A worked run

Here is a fact pattern worked the way you would work it in the exam. The Act is invented. The analysis is not.

The Marine Passenger Services Board cancels Wrenfield Ferries Ltd.'s licence to run a scheduled coastal route. Section 18 of the Marine Passenger Services Act lets the Board cancel a licence where an operator "has abandoned a scheduled route". Wrenfield had suspended sailings for eleven weeks while its only vessel was in refit. The Board held that s 18 was met, found on the evidence that Wrenfield had arranged no alternative sailings, and then used its discretion under s 44 to impose an administrative penalty of $400,000. Section 60 allows an appeal to the Court of Appeal "on a question of law alone".

Open with the presumption. Reasonableness is presumed (Vavilov). Then look for a rebuttal, and s 60 supplies one on the legislative-intent branch. So the framework is Housen, not Vavilov reasonableness — but only for what s 60 actually covers.

Sort each complaint. Take them one at a time.

Does "has abandoned a scheduled route" reach a temporary suspension for refit? This asks what a statutory phrase means, and the answer would govern every other operator. It is a question of law. It falls inside s 60, and the Court of Appeal decides it for itself on a standard of correctness (Housen, as adopted in Vavilov and applied in Bell).

Did Wrenfield arrange alternative sailings? This asks only what happened. It is a pure question of fact, so it is outside an appeal right confined to questions of law. It cannot be raised on the appeal.

The $400,000 penalty. This is the one to sort carefully, and it is where most of the marks are. Do not stop at "it is discretionary, therefore not law." Ask what Wrenfield is actually arguing. Suppose the argument is that s 44 does not authorise a penalty of that kind at all, or that the Board applied the wrong legal test. That is an extricable question of law. It goes to the limits of the Board's statutory power, and Bell confirms that such a question falls inside an appeal right framed in terms of law. If instead the argument is only that $400,000 was too much on these facts, nothing is extricable. That complaint is outside s 60.

Then close the gap. The complaints that fall outside s 60 are not lost. Under Yatar, judicial review remains available for them, and the court's discretion to refuse it is narrow. On judicial review the presumption is not rebutted, so reasonableness applies. You would start with the Board's own reasons and ask whether they are justified against the constraints that apply here. Three are obvious on these facts: the statutory scheme, the evidence before the Board, and the impact on Wrenfield, since a $400,000 penalty on a single-vessel operator is severe.

One thing the standard of review does not answer. Whether Wrenfield was told the penalty was being considered, and got a chance to respond before it was imposed, is a procedural fairness question. It sits outside this framework on judicial review — though note that if it were brought within the s 60 appeal, Abrametz would apply appellate standards to it.

Notice how much of that came from structure rather than knowledge. One statutory provision, read carefully, produced the whole shape of the answer.

What it looks like on the page

Knowing the order is not the same as writing it. Here is the first strand above, in the register an examiner is reading for. Roughly 130 words, and it would take you four or five minutes.

The first issue is the standard of review applicable to the Board's interpretation of s 18. Following Vavilov, reasonableness is presumed on review of the merits of an administrative decision. That presumption is rebutted where the legislature has provided a statutory right of appeal, which indicates an intention that appellate standards apply. Section 60 provides an appeal to the Court of Appeal on a question of law alone. Appellate standards therefore govern: correctness for questions of law, and palpable and overriding error for questions of fact and mixed fact and law (Housen, adopted in Vavilov and applied in Bell). Whether "has abandoned a scheduled route" extends to a temporary suspension is a question of statutory interpretation, and so a question of law. It falls within s 60 and is reviewed for correctness.

Note what that paragraph does. It names the issue, states the rule with authority, identifies the trigger on these facts, and lands a conclusion. No throat-clearing, no history of Dunsmuir, no adjectives. Our guide to writing NCA open-book answers covers the structure at the level of a whole paper.

The exam checklist

1. State the presumption. Reasonableness, per Vavilov. One sentence.
2. Read the Act for an appeal right. If there is one, appellate standards govern everything the appeal right covers — so check its exact terms first.
3. If there is no appeal right, test the rule-of-law categories honestly. If you cannot say what makes the question important to the legal system as a whole, it is not that category. And remember correctness attaches to the question, not the whole decision.
4. Inside an appeal right, sort each complaint into law, fact, or mixed, and look for an extricable legal question before you write anything off as discretionary. Attach a standard to each complaint separately.
5. Check for leftovers. Anything outside a limited appeal right goes to judicial review under Yatar, on reasonableness.
6. For reasonableness, start with the reasons and name the specific constraint your facts engage. "Unreasonable" without a constraint is an assertion, not an analysis.
7. Ask separately whether the process was fair. It is a different question with its own law — unless you are inside a statutory appeal, where Abrametz applies.

The one thing to take away

On the standard of review, the examiner is not testing whether you can recall a word. They are testing whether you can work out which framework governs, and say why. Almost every problem hides that answer in the statute you were given. The answers that score well read the Act for an appeal right before reaching for a case name, and then sort each complaint one at a time instead of giving the whole decision a single label.

If you want context on where Administrative Law sits among the papers, our ranking of the NCA subjects from hardest to easiest may help you plan the order you sit them in.