Tag: disparate impact

Bonkerstown Bypass

The “math is racist” decision is overturned―on narrow grounds that give little guidance for the future

Last week, the Court of Appeal for Ontario issued its decision in Ontario Teacher Candidates’ Council v Ontario (Education), 2023 ONCA 788. The case was the appeal from the notorious “math is racist” decision of the Divisional Court,  Ontario Teacher Candidates’ Council v The Queen, 2021 ONSC 7386, about which I blogged here. The Court of Appeal, contrary to the Divisional Court, holds that the math and pedagogy test (the Math Proficiency Test [MPT]) that aspiring teachers are required to take in Ontario does not discriminate against non-white candidates. But the decision’s scope is very limited. It is based on additional evidence that was not available to the Divisional Court, and while it provides useful guidance on assessing evidence in cases alleging violations of the Charter’s equality guarantee, it neither calls into question the legal framework applicable to such cases (which in fairness, is not the Court of Appeal’s role) nor points out the Divisional Court’s public law mistakes.

Justice Monahan, writing for the unanimous court, sets out the basis for the decision quite succinctly:

The Divisional Court made its findings based on the July 2021 Data, which recorded the results from the approximately 3600 candidates who had attempted the MPT by July 26, 2021. The July 2021 Data disclosed that … White candidates had passed the MPT at a materially higher rate than had racialized candidates.

The July 2021 Data before the Divisional Court provided a preliminary and incomplete basis for assessing the impact of the MPT on entry to the teaching profession. Candidates who had not passed the MPT could rewrite the test an unlimited number of times, and fewer than half of the candidates who would eventually attempt the MPT in 2021 had done so by July 26, 2021. This was an insufficient evidentiary record that failed to discharge the respondents’ onus under s. 15(1) of the Charter.

… December 2021 Data discloses that of the 8350 candidates who attempted the MPT one or more times during 2021, 95% were successful, including 93% of candidates from racialized groups. Moreover, had the MPT been available in 2022, candidates who had not succeeded in 2021 would have had the opportunity to retake the test. Thus, the ultimate disparities in relative success rates between different demographic groups might well be even smaller than the relatively modest differences observed in the December 2021 Data. [6-8]

What matters, according to Justice Monahan, is whether the white and non-white test-takers are able to enter the teaching profession at similar rates, not the outcome of any particular test attempt. There is no evidence that having to retake the test is a material disadvantage; there is no need to wait any significant amount of time before resitting; no harm ― no foul. This is sensible so far as this particular arrangement is concerned, but note how little the decision really matters for the future: in effect, the test is upheld because failure has no meaningful consequences. The test isn’t really much of a test.

Justice Monahan’s comments about statistical evidence in discrimination cases are worth noting, but they too might turn out not to be very consequential. He points out that “the number of candidates who had self-identified as members of racialized groups” in the data considered by the Divisional Court

was quite small, representing a fraction of the total number of MPT test takers in 2021. Moreover, because the absolute numbers were small, relatively small changes in the numbers would have an outsized impact on the success rate (expressed in percentage terms) for racialized candidates. [71; footnote omitted]

Justice Monahan adds that “the Divisional Court did not explain why it was appropriate to draw firm conclusions about the impact of the MPT on racialized teacher candidates in Ontario from such a small sample size”. [71] But that is not fair. The Divisional Court did address this concern. It wrote that demanding more extensive data

is akin to suggesting that more racialized candidates must attempt and fail the MPT to accumulate the data necessary to show a disproportionate impact. The fact that a greater adverse impact could be demonstrated over time does not mean that there is no adverse impact now. While evidence is necessary, it cannot be that a claimant group must wait years before it is in a position to challenge a regulation that it alleges is discriminatory. [Div. Ct., 85]

Of course, the flaw in this reasoning is that it presumes that early trends will simply continue. We know now that they did not, and this might have been anticipated in the circumstances, as the resit attempts allowed groups that were not initially as successful as others to catch up. But then again, the catching-up would presumably not have happened if there had in fact been some deep problem with the test. As I wrote in my post about the Divisional Court’s decision, “[t]here is a logic to” saying that “it cannot be that a claimant group must wait years before it is in a position to challenge a regulation that it alleges is discriminatory”, though “but “that’s because we are used to thinking of constitutional violations that are more concrete and immediate than what spreadsheets capture”.

In other words, a difficult question arises here: how does a court faced with a disparate impact discrimination claim balance the need for sufficiently robust data to support it against the concern about making people wait and suffer discrimination? It is the Court of Appeal that doesn’t explain why it disposes of this question as it does, even as it chides the Divisional Court for its alleged failure to explain itself. The explanation, I suspect, is implicit: facts are stubborn things, as John Adams long ago pointed out, and once the more complete data are admitted into evidence, it would take a peculiarly bloody-minded court to ignore them. But of course this too doesn’t give future litigants and courts a great deal of guidance.

Perhaps this is because the Court of Appeal is simply stuck with a legal framework foisted on it by the Supreme Court, and could not resolve the paradoxes that this framework throws up even if it tried. The difficulties I have just outlined are inherent in allowing statistics to ground claims of unconstitutional discrimination even if no one can offer a plausible hypothesis about how these statistics have come to be. A general invocation of “systemic racism/sexism/discrimination” isn’t an explanation of course ― it simply begs the question. Getting rid of disparate impact claims entirely would of course solve these problems. Short of that, one might at least require a claimant to explain exactly how the system works to produce this specific set of disparate outcomes. Statistical evidence would then be no more than useful supporting evidence, rather than the foundation of the claim, and its weakness would be less of a concern. In any case, these are things for the Supreme Court to think about.

What the Court of Appeal could and should have done better with is the basic question of the issue in this case and, relatedly, the standard of review and the eventual remedy. Justice Monahan describes the issue as “[d]id the Divisional Court err in finding a prima facie breach of s. 15(1) of the Charter … and [d]id the Divisional Court err in holding that any infringement of s. 15(1) caused by the MPT is not justified under s. 1 of the Charter?” [64] But this is ambiguous as to the nature or cause of the alleged “prima facie breach” or, to use more textually grounded language, limitation of the s 15(1) right. Is the claim that s 15(1) is limited and eventually breached by the specific test that was administered to would-be teachers, or to the statutory provisions authorized it?

The distinction doesn’t just matter for the sake of intellectual rigour. Under existing Supreme Court precedent, it makes a difference to the standard of review. If only the actual test that was administered is at issue, it should arguably be reviewed as a discretionary decision, and thus subject to deference under Doré v Barreau du Québec, 2012 SCC 12, [2012] 1 SCR 395 and its progeny. If the statutory provisions themselves were being challenged, then correctness review was in order. The Court of Appeal follows the Divisional Court in applying correctness review, but doesn’t explain whether and why the provisions authorizing a test were being challenged despite the evidence dealing almost exclusively with the test that was actually administered and its outcomes. And of course, if the Court of Appeal had upheld the Divisional Court’s findings, the proper remedy would also have turned on the nature of the claim. As I wrote in my post about the Divisional Court’s decision, I don’t think there was a basis for invalidating the legislative provisions enabling the test, and yet the Divisional Court did just that. It looks like the Court of Appeal could have done the same, and we are none the wiser as to why.

The Court of Appeal’s decision is, obviously, better than the alternative. That alternative, as I wrote in my post, was “desperately sad”:

The Court notes that one problem with the relative paucity of minority teachers is that, at least in the United States, “non-Black teachers had significantly lower expectations of educational attainment for Black students than Black teachers”. [68] Yet its ruling manifests and perpetuates exactly this sort of soft bigotry of low expectations. We can’t expect, it says, “racialized” candidates to succeed on a standardized test twice vetted for sensitivity and bias. … This presumption that letting members of minority groups cut corners is the only way they can succeed is what disregards the constitutional mandate of equality ― not holding them to the same standard as everyone else.

But the survival of this particular test, for no more compelling reason than that it doesn’t ultimately prevent any significant number of people ― of whatever race ― from becoming teachers, isn’t exactly a cause for joy. The contrary outcome would have been bonkers, but the Court of Appeal doesn’t, and in fairness can’t, prevent it from occurring in a future case. It merely bypasses, for now.

The Law of Bonkerstown

Does the constitution mandate the soft bigotry of low expectations?

In a much-noticed decision, Ontario Teacher Candidates’ Council v The Queen, 2021 ONSC 7386, the Ontario Divisional Court has declared unconstitutional a standardized test would-be teachers were required to take in Ontario, as well as provisions of the Ontario College of Teachers Act and regulations made under it that required the test to be administered. The Court found that the test infringed the equality rights of “racialized” candidates under section 15 of the Canadian Charter of Rights and Freedoms, because they (or at least some groups among them) had significantly lower pass rates than white ones. The decision is perverse in itself, as well as showing the perversity of section 15 jurisprudence.

The case can be summarised quite simply. In the face of concerns about falling student performance on mathematics tests, the Ontario legislature introduced a requirement that aspiring teachers themselves pass a standardized test. Regulations were drawn up and a test designed, drawing on a “bank of math questions used for Grade 3, 6, and 9 assessments” for its substantive part and “on Ministry policy documents for the pedagogy component”. [24] Alas, despite two reviews “for bias and sensitivity to equity issues” [25] the results of a trial run and of the first official iteration of the test showed considerable disparities between ethnic and linguistic categories of test-takers. A group of would-be teachers, supported, it is worth noting, by the teachers’ union, challenged the test on the basis that these disparities amounted to unconstitutional discrimination.

In a unanimous opinion by Justices Backhouse and Nishikawa, the Divisional Court agrees. Following Fraser v Canada (Attorney General), 2020 SCC 28, it explains that section 15 rights are limited whenever state action impacts a group defined by a trait it names or an analogous one more severely than a relevantly comparable group, and this disparity “has the effect of reinforcing, exacerbating or perpetuating disadvantage”. [56] Crucially, there is no need for a demonstrated causal connection between the disparity and the disadvantaged group’s characteristics, and no need for all of the disadvantaged group’s members to be similarly affected.

This test is met here by the statistical showing that some non-white groups have done noticeably worse than white test-takers ― a result that the Court ties to the fact that “[t]he education system in Ontario has historically imposed disadvantages on Black and Indigenous students” [64] and that “Black and Indigenous students tend to be over-represented in less challenging, basic and general level courses, and under-represented in advanced courses”. [65] Such students, the Court says, would do better if there were more teachers who shared their background, but there are not ― on the contrary, “[t]wenty-six percent of Ontario students are racialized. However, only thirteen percent of teachers are racialized”. [67] In addition to the statistics, the Court discusses what it describes as “qualitative evidence”, in the form of an affidavit by one candidate, of African origin, who found taking the test difficult. One reason for this was that the test “is taken on a computer”, while the candidate “ha[d] never used a computer or calculator for math and ha[d] never taken a computer-based math test” [89] ― though this candidate was actually successful on the math part of the test, and only struggled with the pedagogy one.

The Court then asks whether this limitation of equality rights is justified in a free and democratic society. It accepts the government’s submission that the test is imposed to improve student performance and is a rational way to do so. Furthermore, the issue involves balancing would-be teachers’ rights and the needs of the students, which points to “a highly deferential approach” [132] to deciding whether the test is a minimally impairing way of achieving the government’s purposes. Yet the Court still finds that the test is not justified. This is, crucially, because the government would do as well ― or, the Court strongly suggests, better ― if the test were replaced with a requirement that aspiring teachers be required to take a math course as part of their mandatory education. The Court’s key finding is that this

would be significantly less impairing of equality rights. There is a plethora of evidence in the record highlighting the disproportionate impacts of standardized testing based on race … In contrast, there is no evidence to suggest that these negative diversity impacts would exist in the context of a mandatory math course (i.e. that racialized teacher candidates might disproportionately fail these math courses), and we do not think such an inference can be made on the basis of logic or common sense. [142]

This way lies madness. The Divisional Court is, of course, bound by the Supreme Court’s test at the section 15 stage of the analysis, and for the most part it should not be faulted for the way in which it applies it. What its application of the test makes clear, though, is that this test is, to use a technical term, bonkers. It compels the conclusion that constitutional rights are limited ― or, as the Court puts it, prima facie breached ― by what could, for all we know, be statistical artefacts. (One point that arises in the Court’s discussion is whether data from one trial and one official test are sufficient to support its conclusions. There is a logic to the Court’s response that “a claimant group must wait years before it is in a position to challenge a regulation that it alleges is discriminatory”, [85] but that’s because we are used to thinking of constitutional violations that are more concrete and immediate than what spreadsheets capture.) References to historical discrimination and current statistical disparities are only for colour; they do are not meant to have any actual explanatory power. They cannot, for instance, account for the fact that, on the trial edition of the test, the disparity between candidates of “Caribbean” and “African” backgrounds was almost as great as that between “Caribbean” and “European” candidates. A potential constitutional violation is made out by a showing no one needs to and, quite possibly, can explain. This is a surrender to, indeed an embrace of, irrationality masquerading legal doctrine.

Meanwhile, the Divisional Court’s approach to section 1 is perverse in its own right, though it is also consistent with the underlying logic of the Supreme Court’s approach. The Court says that the province should rely on courses in mathematics rather than tests to ensure that teachers are competent, because there is no evidence of disparate effect of a course requirement. For one thing, note that the Court doesn’t say there is evidence of equal outcomes. It’s just that nobody seems to have studied the matter. And consider: if the course uses rigorous assessment of the same kinds of skills that the test measured, why would the outcomes be different? Alternatively, how will we know that the reason for the course requirement not yielding disparate outcomes won’t be a lack of rigorous assessment? I’ve taught enough at universities to know that this is not exactly a far-fetched hypothesis.

Here is the desperately sad thing about the Court’s reasoning. The Court notes that one problem with the relative paucity of minority teachers is that, at least in the United States, “non-Black teachers had significantly lower expectations of educational attainment for Black students than Black teachers”. [68] Yet its ruling manifests and perpetuates exactly this sort of soft bigotry of low expectations. We can’t expect, it says, “racialized” candidates to succeed on a standardized test twice vetted for sensitivity and bias. The “qualitative evidence” on which the Court relies is telling in this regard. It is, apparently, too much to ask of a teacher, or at any rate, on the Court’s logical, a “racialized” teacher, in the 21st century to be able to do a test on a computer. We shouldn’t care that a future teacher cannot answer multiple-choice questions about pedagogy, because hey, he is “uniquely qualified to broaden students’ awareness and understanding of Franco-African culture and cross-cultural world views”. [91] This presumption that letting members of minority groups cut corners is the only way they can succeed is what disregards the constitutional mandate of equality ― not holding them to the same standard as everyone else.

And yes, of course, it’s not impossible that the standard is set in an insidiously biased way. Needless to say, a standard vetted, twice, for precisely this potential problem by people genuinely concerned about this possibility is unlikely to have this problem. Still, perhaps. But then it should be possible to demonstrate that this is so. As Justices Brown and Rowe say in their dissent in Fraser, one should not assume that a statistical disparity in outcomes between groups or a correlation between disadvantage and membership in a group equals causation:

where one is dealing with complex social and economic considerations … one can readily assume that there are many factors involved, some of which will give rise to causation while others will simply be the result of coincidence (that is, caused by independent factors). [180; emphasis in the original]

Let me end this part of the past by quoting Chad Williams, writing for Fair for All:

the fundamental assumption of critical race theory is that disparities between racial groups are in themselves evidence of racism. This assumption is faulty because it fails to allow or account for the numerous reasons other than racism that racial groups might have different outcomes. At the population level, these plausible reasons include age, geography, and, of course, culture. Many analysts of group differences are trapped in the false binary that these differences must result from either racism or genetic differences. However, Thomas Sowell and others have made compelling arguments about why this simply is not true. This is not to say that past disparate treatment has had no effect on current disparate outcomes, but it certainly cannot be the only cause for these disparities. Accordingly, a policy that begins from the faulty assumption that disparities equal racism is bound to fail as a bulwark against what actual racism still exists today.

This policy is the law the Supreme Court’s created in Fraser, and the Divisional Court’s decision in Candidates’ Council illustrates the tragic consequences it is bound to have.

Before addressing the question “and now what?”, let me point out two things about this case that I do not understand. One has to do with what exactly the Court is deciding. Beyond explaining that a recent amendment to the College of Teachers Act required the test to be set up and regulations under the Act gave effect to the requirement, the Court’s reasons focus entirely on the test. The evidence, such as it was, had to do with the results of different groups of candidates on the particular test they were administered, and the experience of some individual candidates. Yet the court not only invalidates the test, but also the statutory provision and the regulation pursuant to which it was created. It doesn’t at all explain why. Under the heading “What Is the Appropriate Remedy?” it simply states that “[a] declaration” to this effect “will issue”. [161] The province can’t create a new test even if thinks that it might come up with one that will not produce the disparities this one did. That doesn’t seem right to me, and the absence of any explanation does not speak well of the Court’s commitment to transparency and to judicial craft.

Relatedly, I do not understand the Court’s approach to the standard of review. The Court deals with it briskly, citing  Canada (Minister of Citizenship and Immigration) v Vavilov, 2019 SCC 65, for the proposition that “constitutional questions are reviewed on a standard of correctness” [50]. But that seems too quick. Vavilov makes a

distinction between cases in which it is alleged that the effect of the administrative decision being reviewed is to unjustifiably limit [Charter] rights … and those in which the issue on review is whether a provision of the decision maker’s enabling statute violates the Charter. [57]

The former type of cases is not, ostensibly, decided on the correctness standard at all, but following the framework set out in Doré v Barreau du Québec, 2012 SCC 12, [2012] 1 SCR 395 and refined in a number of subsequent cases. Only the latter attract the correctness standard. Co-blogger Mark Mancini has argued that this is incompatible with Vavilov‘s overall thrust, and so have I, but that’s how it is until the Supreme Court says otherwise. Now, as I’ve alluded to just above, Candidates’ Council seems to involve both types of issues. The design of the test is a discretionary administrative decision. It’s not prescribed by statute or regulation: regulation 3 of the Proficiency in Mathematics Regulations, made under the College of Teachers Act, simply provides that “[T]he Education Quality and Accountability Office shall develop English and French versions of a mathematics test”. The test, then, should in principle be assessed under the Doré framework, on a reasonableness standard. Meanwhile, the Regulations and the enabling section of the Act are obviously the kind of provision whose constitutionality falls to be addressed on the correctness standard under Vavilov.

Does that make sense? No, not really. I would even say, none at all. But that’s what Vavilov pretty clearly seems to require. It is puzzling that the Court does not even attempt to engage with its demands. This isn’t even a matter of fudging the law to get to a desired outcome ― as we have seen, the Court makes a show of being deferential while ostensibly applying the correctness standard, and strikes down the test anyway. It could easily have reached the same result under Doré. I suppose the absurdity of the Vavilov/Doré cohabitation was simply too much to deal with.

So what to do about this terrible decision? Unsurprisingly given both the Zeitgeist and the egregious nature of this decision, the n-word ― “notwithstanding” ― has been thrown around since the decision came out. More than on other occasions, I understand the impulse. But, as before, I remain of the view that it must be resisted. While using the “notwithstanding clause” to undo this particular decision may be a short-term fix, it is not a genuine solution to the underlying problem. The invocation of s 33 of the Charter will have to be renewed every five years. A legislature that takes a different view of either the substantive issues or indeed the propriety of using the notwithstanding clause itself need not do so. More importantly, using the notwithstanding clause to resurrect this particular policy does nothing to prevent other judgments, striking down other policies, from applying the same misbegotten principles as the Divisional Court applied here. Are the supporters of the notwithstanding clause saying that it should be used, René Lévesque-style, to insulate all law from judicial review? They’ve always been angry at me for suggesting that that that’s where their positions naturally led.

The immediate solution, of course, is for the province to appeal. One should hope that the Court of Appeal, and the Supreme Court if need be, will restore sanity, if only on the section 1 issue, on which Fraser does not control. Otherwise, the appropriate solution ― and it is probably well worth pursuing regardless of the outcome of the appeal ― is an amendment to section 15 of the Charter. At a minimum, it should endorse the dissenting position in Fraser, requiring a showing of causation for statistical disparities to be regarded as limiting equality rights. More radically, the amendment could excluding disparate impact discrimination from section 15’s scope. My tentative opinion is that this is the better, cleaner option ― but I am open to persuasion. Either way, an amendment is both a more permanent ― not to mention transparent and honest ― way of avoiding similar disasters in the future than using the notwithstanding clause to revive this particular policy, and a more modest one than a blanket invocation of the notwithstanding clause to prevent section 15 challenges at all would be.