Introduction
The precautionary principle holds that the state may limit citizens’ rights in order to proactively avert serious harm even if there is factual uncertainty about that harm’s severity, the likelihood it will materialize or the efficacy of policies proposed to mitigate it.Footnote 1 The principle, traditionally invoked in environmental law,Footnote 2 became especially salient during the COVID-19 pandemic when governments worldwide imposed restrictive measures that curtailed individual freedoms to minimize the virus’s threat to public health.Footnote 3 These measures included lockdowns, limits on the sizes of public gatherings, travel restrictions, physical distancing requirements and vaccine and mask mandates. At first, the utility of these measures was unknown because so much about the virus was unknown. How was it transmitted? How deadly was it? Who was most vulnerable? Would the restrictions be effective and were they necessary? In the face of scientific uncertainty, governments could wait for conclusive evidence about these matters to emerge before acting, thereby risking catastrophic consequences on public health. Usually, however, they chose to act proactively even while the scientific credentials of their policies remained unclear.
Consider the situation in Canada as a case in point. Many restrictive measures imposed during the pandemic in the name of precaution were alleged to infringe the Canadian Charter of Rights and Freedoms. Footnote 4 Claimants challenged public gathering restrictions as limiting religious congregations contrary to the right to religious freedom guaranteed under Section 2(a) of the Charter. Footnote 5 Restrictions on travel that prohibited interprovincial movement were said to limit the right to free mobility protected by Section 6.Footnote 6 Lockdowns allegedly limited the right to liberty and, given the potentially damaging psychological effects of isolation, one’s right to security of the person under Section 7.Footnote 7 It was also argued that vaccination mandates limited the right to equality under Section 15 of the Charter by discriminating on the basis of vaccination status.Footnote 8 If any of these arguments were to succeed, it would fall to the state to show that the limit on Charter rights was reasonable under Section 1, the Charter’s justification provision.Footnote 9 According to the test for justification set out in the Supreme Court of Canada’s well-known R. v. Oakes judgment, the impugned law must have a pressing and substantial objective and the limit on the Charter right must (i) be rationally connected to that objective; (ii) impair the right as little as possible and (iii) have overall benefits that exceed its overall costs.Footnote 10
However, what if the state is unable to adduce evidence in support of its position on justification because the evidence is inconclusive, or if the perceived harm it seeks to prevent is not amenable to proof? Two options present themselves. First, we might hold that it is never possible to justify a law enacted in the context of evidentiary uncertainty even if the law aims at preventing an immanent yet not easily quantifiable risk of harm. For example, if a law’s efficacy is unknown, it would be difficult to establish that less drastic means would be less effective in achieving its goal or that the law’s benefits exceed its costs. A second option is to relax the evidentiary requirements needed to justify laws enacted in factual uncertainty. However, without more, this approach allows potentially boundless judicial deference to governments, which can preclude any meaningful review of government action or, as we discuss later, create opportunities for courts to decide constitutional disputes arbitrarily. Since both options are unsatisfactory, the result is an apparent dilemma.Footnote 11 The search for an appropriate framework that avoids these problems has engendered only limited debate among comparative constitutional theorists,Footnote 12 and the scholarly discussion has largely gone unaddressed by constitutional courts.
It might be thought that the precautionary principle supplies a ready answer to the dilemma. However, unless the principle is suitably elaborated, it quickly submerges us in it. As an inherently permissive principle, precaution allows us to circumvent the first horn by eschewing the need for evidentiary certainty as a condition for justifying proactive measures. However, if it gives no further guidance in setting out ascertainable criteria for how the state must exercise this prerogative, it impales us on the second horn by licensing unstructured curial deference to the state and its choices about when to limit constitutional rights when critical facts are unknowable.Footnote 13
In this article, we defend a role for precaution within proportionality analysis that aims to chart a path out of the dilemma. We argue that notions of risk analysis already embedded in long-accepted legal doctrines can inform proportionality in a manner that suitably controls the permission that the precautionary principle gives governments to justify rights-limiting laws in situations of factual uncertainty. Our contribution advances legal knowledge in this area in at least three respects. First, unlike approaches to this problem offered by other comparative constitutional commentators, our approach is appropriately modest in the sense of avoiding excessive revision of well-settled tenets of proportionality reasoning. Second, we take up a universal perspective that considers how to incorporate precaution within proportionality generally without concentrating singularly on limits on specific rights, most notably the right to religious freedom.Footnote 14
Our third contribution emerges from our focus in this article on Canadian jurisprudence, which, as alluded to already, functions as our comparative case study. We by no means suggest it is only by attending to the Canadian experience that we can best comprehend the relationship between proportionality and precaution. Indeed, we acknowledge that the Oakes test has its precursors in nineteenth-century German proportionality analysis that gained subsequent worldwide ascendance.Footnote 15 As well, to date, most of the existing comparative writing with which this thesis engages emanates from scholars of European constitutionalism. However, we submit that Canada is nevertheless instructive for comparative constitutional scholars because its approach to proportionality in human rights law cases has been highly influential in jurisdictions elsewhere, especially in common law countries that have expressly adopted the Oakes test into their own constitutional frameworks.Footnote 16 As put by Mark Tushnet, the perspicacity of Oakes makes it ‘readily adoptable by other courts’.Footnote 17 The Supreme Court of the United Kingdom has gone so far as to call the contemporary Canadian iteration ‘the clearest and most influential judicial analysis of proportionality within the common law tradition of legal reasoning’.Footnote 18
However, ironically, what truly recommends the Canadian experience is that it provides a cautionary tale for other countries on this occasion. For, it neatly illustrates the pitfall of invoking precaution within proportionality, that is, entanglement between the Scylla of government inaction to forestall harm and the Charybdis of freewheeling judicial deference. The Supreme Court of Canada has over time become entrapped in this dilemma. Despite initially enunciating a rigorous evidentiary standard for proportionality in Oakes, the Court has steadily retreated from this standard and permitted the state to rely on a ‘reasoned apprehension of harm’, or resort to a priori standards of logic and common sense, to justify rights-limiting laws in factually ambiguous contexts. The predictable difficulty has been that, lacking any more detailed structural guidelines for the exercise of these permissions, the softened, deferential evidentiary stance has been applied inconsistently and has led to imprecise and subjective application of the Oakes proportionality framework.
We suggest that other areas of Canadian law already contain well-established methods of managing risk of harm that are adaptable to the constitutional context. We argue that, in place of the amorphous invocations of apriorism in proportionality reasoning, a structured assessment of the risk and gravity of harm at each step of the Oakes test in accordance with risk analysis addresses the conditions for justifying laws amidst epistemic uncertainty in a more objective and orderly fashion.
Putting this together, our analysis unfolds as follows. In Part 2, we explain how the Supreme Court departed from a strict standard demanding proof of demonstrable evidence under Oakes and adopted a more permissive one when confronted with factual uncertainty. This has invited the unanswered critique that proportionality under Section 1 of the Charter has become a medium for unprincipled decision-making that turns on arbitrary degrees of deference to legislatures. We then analyze lower court cases concerning the justification of limits on Charter rights to protect public health during the COVID-19 pandemic. These cases offer an opportunity to revisit the well-worn critique of how the Supreme Court has applied the attenuated evidentiary standard from a novel contemporary perspective. We identify familiar problems in these decisions stemming from the absence of formal guidelines for operationalizing deference and precaution within proportionality. We then turn to developing such guidelines in Part 3. In our view, when applying Section 1 of the Charter, courts must assess the level of risk and the gravity of the harm that a law is said to combat. We analyze how the approach operates in Canadian tort and criminal law and recommend integrating a similar analysis into each stage of the Oakes test. We maintain that risk assessment gives more definite content to the epistemic standards of reasonableness, logic and common sense throughout the Oakes test. We emphasize the modesty of our proposal, which entails only minor adjustments to the test, and contend that this gives it an advantage over alternative proposals advanced in the comparative scholarly literature, which mandate a complete overhaul of settled proportionality principles. We therefore submit that our approach is ripe for application in other countries whose courts have drawn guidance from Oakes when applying proportionality in the contexts of factual uncertainty.
Perils of relaxed evidentiary standards under Oakes
Before the pandemic, Canadian courts invoked precaution to justify government action infrequently. However, proportionality analysis amid inconclusive evidence is not new. In what follows, we explain the Supreme Court of Canada’s early adoption of a flexible evidentiary standard when applying the Oakes test in situations of evidentiary uncertainty. We then outline the shortcomings of this approach and concerns about how it invites excessive judicial subjectivity. We then show how those concerns have resurfaced in recent cases where other Canadian courts have upheld measures to stop the spread of COVID-19 using the precautionary principle.
From empiricism to rationalism under the Oakes test
In Oakes, the Court originally set out a ‘stringent standard of justification’ whereby limits on Charter rights would be justified only if ‘exceptional criteria’ were met.Footnote 19 Charter rights would be privileged ‘in the normal course’ and limits treated as aberrant deviations requiring a special defense.Footnote 20 To that end, the state had to adduce ‘cogent and persuasive evidence’ that would ‘make clear’ the consequences of imposing or not imposing a given measure to achieve a given objective.Footnote 21 Thus, the epistemic orientation of Oakes was decidedly empirical.Footnote 22
Almost immediately after Oakes, the Court relaxed this empirical attitude and shifted toward a more deferential standard.Footnote 23 This was especially so where it considered the justification of laws passed in the contexts of evidentiary uncertainty. In Irwin Toy, a toy manufacturer challenged a law that prohibited companies from targeting advertisements to children under 13 years old as unjustifiably limiting their expressive freedom. The evidence supporting the use of this age range was unsettled, and the age range potentially captured audiences that did not need protection from commercial manipulation. Without clear evidence, a strict reading of Oakes would suggest that the law could not be justified. However, the Court stated that judges should be slow to second-guess legislators who have made a reasonable assessment of how to draw regulatory distinctions in the face of unsettled evidence. Otherwise, they ‘would only be substitut[ing] one estimate for another’.Footnote 24 Hence, where evidence is inconclusive, courts must afford a measure of deference to the expertise of legislators.Footnote 25
The Court expanded on these claims in Keegstra. Footnote 26 A teacher who made anti-Semitic comments to his students argued that the crime of hate speech did not proportionately limit the right to free expression because the state was unable to prove that hateful statements harm anyone. The Court admitted that it is difficult to connect hate speech to harmful downstream effects but held that where harm was incapable of being proven through evidence, the state may instead rely on a priori standards of common sense, reason and logic. There was a logical inference that hate propaganda would cause harm that the government had a legitimate public interest in preventing. This rationalist approach was affirmed in Butler, where an adult video store owner challenged obscenity crimes as unjustifiably infringing his expressive freedom. The evidence could not establish a causal link between obscenity prohibitions and the government’s stated objective of protecting women and children from the harms of exposure to dehumanizing material. However, the Court accepted that the state could nevertheless criminalize obscene materials if it has a reasonable basis to believe that such a link existed based on a ‘reasoned apprehension of harm’.Footnote 27
The reasoned apprehension of harm standard for justifying laws, as well as logic, reason and common sense, were thus substituted for the standard of cogent and persuasive evidence previously established in Oakes. They eventually coalesced into a rationalist epistemic orientation designed to effectuate judicial deference to legislative decision-making in the contexts of uncertainty. They would then govern proportionality analysis in such contexts going forward.
In Thomson Newspapers, a news organization challenged a law banning the publication of polling data in the lead up to an election. The government argued that the law was necessary to prevent the harm that could ensue if voters were unduly influenced by inaccurate polls before election day, but it did not provide conclusive evidence to prove this problem existed. It instead argued it was acting upon a reasoned apprehension that such a harm could occur. The Court held that since the alleged harm was not in the everyday knowledge or experience of Canadians, the government could not rely on common sense or logic to fill in evidentiary gaps.Footnote 28 Yet, these a priori standards were satisfied in Harper, which involved the justification of spending limits on third-party advertisers to prevent private interests from overwhelming political airwaves during elections. There was no evidence that electoral fairness would be compromised in the absence of spending limits, but the state could rely on a reasoned apprehension of that harm, and, given the seriousness of the harm and uncertainties surrounding how it could materialize, deference was warranted.Footnote 29 Similarly, in Bryan, the Court upheld a law prohibiting news agencies from reporting on early electoral results to ensure an equal distribution of information among voters. Without evidence demonstrating inequities in voting behavior, the Court accepted that, as a matter of logic and common sense, premature release of electoral results would compromise public confidence in the electoral system.Footnote 30 A ‘deferential approach’ to justification was necessary, and ‘traditional forms of evidence (or ideas about their sufficiency) may be unavailable in a given case and that to require such evidence in those circumstances would be inappropriate’.Footnote 31
How rationalism abets arbitrary judicial decision-making
What is striking about these cases succeeding Oakes is how the Court struggled to find a consistent home for the relaxed a priori standards within the Oakes proportionality test. For instance, in Butler, the Court acceded to a reasoned apprehension of harm of obscenity to decide that criminalization was rationally connected to that harm. This approach was reaffirmed in Sharpe, where, in the absence of adequate proof, the Court recognized a rational connection between the prohibition of child pornography and the reduction of harm to children.Footnote 32 In the election law cases of Thomson Newspapers, Harper and Bryan, by contrast, standards of reason, logic and common sense were primarily raised to inform the ‘context’ in which the Oakes test applies, with the Court making only cursory reference to them at specific stages of the analysis. More recently, in Whatcott, the Court invoked the reasoned apprehension of harm of hate speech only at the minimal impairment stage to assess whether certain provisions of Saskatchewan’s human rights code went farther than necessary to reduce the harmful effects of discriminatory expression.Footnote 33
The unstable, roving character of rationalism’s appearances at various stages of the Oakes proportionality analysis invites a critique that the Court has merely improvised the methodology for invoking it to reach a desired result in the case at hand. That is, if rationalist standards have no fixed place under Oakes, properly applying everywhere and nowhere all at once, they can be applied at whatever stage they are needed to induce curial deference to government or they can be omitted entirely if the Court is unwilling to defer. In addition, once a court decides to defer, it is unclear how much deference it should afford at each stage of the justification analysis. For example, even if there is no question that the legislature must be shown a high degree of deference, it is unclear how this differs from low deference when the deference scale itself remains vague within the steps of proportionality analysis. The imprecision of the role for rationalist standards under Oakes provides a fertile environment in which judicial arbitrariness and subjective, results-driven decision-making can flourish, even though deference is ostensibly introduced to curb judicial subjectivity.
This critique echoes one that has been voiced in various forms by others in the years since Oakes-style empiricism waned. For example, Guy Davidov depicts rationalism’s liability to engender selective judicial deference in terms of a paradox.Footnote 34 We saw that the Court in Irwin Toy initially abandoned the empiricism of Oakes when confronted with uncertainty because it acknowledged that somebody had to make a nonevidence-based judgment about how best to achieve an important public objective, and this judgment ought to be made by democratically accountable legislators rather than unelected judges.Footnote 35 However, if there are no tight strictures for how rationalist standards are incorporated into proportionality, courts are left with wide discretion about when and how to invoke them. As Davidov puts the point, if the strictures remain vague, the concept of deference ‘undermines its own justification’Footnote 36 by becoming ‘a perfect tool’ for judges: ‘powerful when they want it, meaningless when they don’t. It creates an illusion of limiting subjectivity, but at the same time gives the courts unlimited discretion to achieve any desired result’.Footnote 37
Davidov’s critique dovetails with concerns about not just where within Oakes rationalistic deference is triggered, if it is triggered at all, but also how, once triggered, a rationalist approach should be applied and the corresponding degree of deference to which the legislature is entitled. It is not clear how courts are to go about determining how exacting they must be in scrutinizing whether the government’s apprehension of some societal harm is reasonable or whether logic and common sense dictate that the government’s means are rationally connected to its goal to address that harm. For instance, it is difficult to discern why in Thomson Newspapers the apprehension of harm that inaccurate polls could unduly influence the electorate was ‘speculative’, whereas in Bryan, the apprehension of the harm that early electoral returns from one side of Canada would unduly influence voters on the other side of the country was not. Such a lack of clarity allows courts to defer selectively on an outcome-driven basis even if it is possible to settle the question of whether the application of a given stage of Oakes calls for rationalism.Footnote 38
In response to these critiques, one might suggest that the Court’s decision in RJR-MacDonald spawned a line of case law that seemingly attenuates the shift toward rationalism under Oakes and reintroduces some empiricism to lessen the judicial subjectivity inherent in rationalism’s deferential posture. RJR-MacDonald concerned the justification of a federal law limiting free expression by requiring tobacco companies to place health risk warnings on cigarette advertisements. All the judges agreed that the lack of scientific certainty about the root causes of tobacco consumption and the efficacy of warnings warranted some level of deference to Parliament at the rational connection and minimal impairment stages of Oakes. The dissent saw it as appropriate to defer to Parliament’s position that a full rather than partial ban on warning-free advertisements minimally impaired free expression, writing that ‘it is not the role of this Court to substitute its opinion for that of Parliament concerning the ideal legislative solution to this complex and wide-ranging social problem’.Footnote 39 The majority, however, refused to defer at the minimal impairment stage, stating that Parliament gave no argument for why a total ban was necessary. Even in the face of factual uncertainty, ‘Parliament does not have the right to determine unilaterally the limits of its intrusion on the rights and freedoms guaranteed by the Charter’,Footnote 40 and if it did then ‘ad hoc judicial discretion’ on the fitting level of deference would override the need for ‘reasoned demonstration’ that limits on rights are proportionate.Footnote 41
The majority’s view was reiterated in Chaoulli, where the Court concluded that a Quebec law prohibiting people from purchasing private medical insurance was not justified. At the minimal impairment stage, the Court appeared to accept that because the evidence was inconclusive as to whether private insurance had a negative effect on public healthcare, the government was entitled to advance arguments based on logic and common sense that it would. However, when witnesses who testified in support of these arguments could not ‘cite specific facts in support of their conclusions’, the Court refused to defer. Thus, despite the factual uncertainty, the Court insisted on evidence and proof.Footnote 42 It is difficult to reconcile this approach with the reasoned apprehension of harm framework that featured prominently in the free expression cases discussed above.Footnote 43 Similarly, in Carter, the Court found that the criminalization of medical assistance in dying was not justified despite conflicting evidence on whether decriminalization would expose vulnerable people to abuse. The Court declined to defer to the government’s asserted apprehension of harm at the minimal impairment stage and said that justification demanded ‘demonstration, not intuition’, which was to be resolved by evidence.Footnote 44 Again, it is unclear why the Court refused to frame its analysis by evaluating whether the government’s apprehension of harm was reasonable against the backdrop of evidentiary ambiguity.
While RJR-MacDonald, Chaoulli and Carter might be taken to represent a course correction of sorts, they only show that, even if there is consensus that courts may rely on rationalist standards of logic and common sense in the absence of clear and cogent evidence, and even if it is clear where these standards operate within the Oakes test, there remains a problem of specifying what counts as ‘reasoned demonstration’ when applying the standards. This benchmark presumably requires evidence that goes beyond a bald governmental assertion of harm or its policies’ efficacy yet need not rise as high as clarity and cogency. The concern is that without precise guidelines for defensibly locating points within this range, the threat of ‘ad hoc judicial discretion’ about when and how to defer still looms large.
The precautionary principle and COVID-19: aggravating an old worry
In recent years, however, the Supreme Court has rarely discussed how proportionality applies when facts are uncertain. However, this was one of the many legal problems raised by the COVID-19 pandemic, which, as we demonstrate presently, resurfaced timeworn worries about softened evidentiary standards under Oakes. Courts have deferred to elected representatives by permitting them to limit Charter rights on a precautionary basis to stop the spread of COVID-19 even when definitive scientific knowledge about the virus or the best public health policies was unavailable. However, the role of the precautionary principle within Oakes has not been clarified, nor has the precise amount of deference courts must show when the state asserts it.
In Taylor, the Supreme Court of Newfoundland and Labrador held that executive orders restricting entry into the province justifiably limited the Section 6(1) Charter right to mobility. When considering whether less intrusive measures, such as requiring temporary self-isolation after entry, would adequately protect public health, the court deferred. It accepted expert testimony in support of the province’s position while remarking that ‘[i]n the context of the COVID-19 pandemic … the public health response is to err on the side of caution until further confirmatory evidence becomes available; the precautionary principle’.Footnote 45
For similar reasons, the Manitoba Court of Appeal in Gateway Bible held that public gathering restrictions prohibiting religious congregations minimally impaired freedom of religion. The scientific evidence could not establish with certainty that COVID-19 could not be transmitted outdoors. The court deferred to the government’s view that the gathering restrictions had to apply to outdoor congregations to protect public health, stating that the government ‘cannot be faulted for taking a precautionary approach of limiting gathering sizes given the difficulty of enforcing physical distancing and mask wearing while outdoor at public places’.Footnote 46
In Grandel, the Court of King’s Bench for Saskatchewan held that public gathering restrictions preventing citizens from protesting antiviral measures in large groups justifiably limited free expression. Whereas the Taylor and Gateway Bible courts invoked precaution only at the minimal impairment stage of Oakes, the Grandel court prefaced its entire Section 1 analysis with the claim that the precautionary principle provided relevant context for the whole Oakes proportionality test. The government ‘could not wait for scientific certainty in order to act in a situation where catastrophic loss of life was at risk’, and the court agreed that ‘when an activity raises threats of harm to human health or the environment, precautionary measures should be taken even if some cause and effect relationships are not fully established scientifically’.Footnote 47 It went on to find that there was a basis in reason and logic that the gathering restrictions were rationally connected to protecting public health and that the government should not be held to a standard of ‘scientific certainty’ when determining whether the restrictions were minimally impairing.Footnote 48
The Ontario Court of Appeal followed Grandel in Trinity Bible Chapel by affirming the use of precaution to justify restrictions on both outdoor and indoor gatherings as minimally impairing freedom of expression. The motion judge maintained there was some, albeit inconclusive, evidence of the risk of transmitting COVID-19 outdoors, and ‘where there are threats of serious, irreversible damage, lack of full scientific certainty is not a reason to postpone harm reduction strategies’.Footnote 49 The Court of Appeal stated that the government is entitled to rely on a reasoned apprehension of harm when determining whether alternative policies would achieve its objectives and is owed deference when the facts are uncertain.Footnote 50 It noted that the minimal impairment analysis ‘still requires an evidentiary basis to show why a measure is a reasonable means of achieving a pressing and substantial objective … the precautionary principle helps inform what it means to rely on a reasoned apprehension of harm where scientific certainty is not possible’.Footnote 51
To support the reasoned apprehension of harm test, the Trinity Bible Chapel court referred to the Supreme Court’s election cases, such as Thomson Newspapers and Harper, which established that factual uncertainty must be considered as the context for the Section 1 Charter analysis. It upheld the motion judge’s approach of applying rationalist standards of logic and common sense under the minimal impairment step of Oakes. Footnote 52 However, the claim, also expressed in Grandel, that precaution informs the whole context of Oakes leaves unanswered whether minimal impairment, rational connection or other stages of Oakes, or indeed whether all the stages, should be the proper site of precaution. And again, if precaution’s placement is amorphous, it can be applied everywhere and whenever judges feel they must defer to the state. They invariably have deferred in the case of the coronavirus restrictions imposed in Canada. On the other hand, if judges feel they must not defer, the claim that precaution is contextual enables courts to decline to apply it anywhere specifically within Oakes.
It might be argued that in many of the COVID-19 cases discussed above, the precautionary principle and attendant rationalist epistemic standards have been routinely applied under the minimal impairment stage, which suggests that this is their uncontroverted home.Footnote 53 Yet if evidentiary uncertainty must pervade the entire context of Section 1, it is unclear why they should be confined to minimal impairment. In fact, they were once thought to be most germane to rational connection in decisions such as Keegstra and Butler. And even if it is undisputed that precaution informs minimal impairment, there remains indeterminacy as to how it does so. The Trinity Bible Chapel court emphasized that even when precaution informs the Oakes inquiry into less drastic alternatives, the state must be held to some standard of proof. The state is not entitled to automatic deference by asserting that it has apprehended harm and that its policies are needed to prevent the harm. However, we saw that the requisite standard of proof within the realm of rationalism is ill-defined. Wittingly or not, courts may leverage this vagueness to hold that the standard is met and automatically attain a desired outcome. Conversely, as in Chaoulli and Carter, a judge may refuse to defer if he/she believes it is not met and desires a different outcome. The lack of precise delineation perpetuates the problem of excessive judicial discretion. That none of the courts across Canada have pushed back against governments’ restrictive measures during the pandemic raises a worry – or rather aggravates an old worry – that precaution is merely a mechanism to prompt judicial deference to elected authorities.
A risk-based approach to proportionality
We now turn to possible solutions to the problems that arise when integrating precaution into proportionality. We first discuss recommendations made by others in the comparative constitutional literature, expose their pitfalls and draw out some lessons to learn. Taking these lessons seriously, we then explore how risk management may justify laws that limit rights on a precautionary basis when evidence is uncertain.Footnote 54 On this approach, the level of judicial deference to the legislature should reflect the gravity of the harm the state seeks to mitigate and the likelihood that the harm will materialize. We explain how Canadian courts have adopted a similar approach in tort and criminal laws and outline how it can be incorporated into each stage of the Oakes test under Section 1 of the Charter. We submit that risk analysis improves on extant scholarly recommendations because it reflects a modest adjustment to proportionality analysis that helps rid the analysis of judicial subjectivity.
Reforming proportionality: recommendations, pitfalls and takeaways
Just as others have identified the paradox of ad hoc judicial discretion stemming from the rationalist turn in proportionality, some have proposed solutions to it. First, it has been suggested that in the category of cases featuring factual uncertainty, the burden of proof (including the burden of persuasion and the evidentiary burden) should be on the claimant to demonstrate not only that a law limits a right, but that the limit is disproportionate. Deference to government is built into the burden shift, which supposedly suppresses selective, results-oriented deference by taking the choice about when and how to defer out of the courts’ hands.Footnote 55
However, this recommendation would completely revise the usual structure of constitutional adjudication and, in the process, erect barriers to accessing constitutional justice that threaten to make allegedly invalid laws immune to challenge. In Canada, the onus of justification under Section 1 of the Charter has always been on the government, and the Supreme Court has flatly rejected suggestions to the contrary.Footnote 56 More broadly, reversing the justificatory burden of proof devalues the constitutional protection of rights, and, since litigation is resource-intensive, reinforces the claimant’s disadvantage against better-resourced governmental entities.Footnote 57
A second recommendation is to modify the final stage of the Oakes test (balancing the overall harms and benefits of an impugned law) in situations of evidentiary uncertainty by relaxing rather than reversing the government’s burden of proof. Legislation would only be disproportionate if its deleterious effects are ‘clearly’ or ‘grossly’ disproportionate to its salutary ones, as opposed to being disproportionate simpliciter. Footnote 58 Deference to the state is taken out of judges’ discretionary hands and codified in the relaxed justificatory burden, but the judicial function of holding the government constitutionally accountable is still preserved in attenuated form. This recommendation revises the Oakes test less than the first, as it draws upon a defined legal framework that is already embedded under Section 7 of the Charter, which protects against laws that deprive persons of life, liberty or security of the person in a grossly disproportionate manner.Footnote 59 Gross disproportionately occurs not when a law’s costs simply outweigh its benefits, but when they are ‘totally out of sync’ with the law’s objectives.Footnote 60
However, even if it is tolerably revisionary, this second recommendation is silent on how courts should apply the precautionary principle at earlier stages of the Oakes test. These stages are arguably more important given that the Supreme Court of Canada has rarely struck down a law at this final stage after it was persuaded that the government had satisfied all the others.Footnote 61 Introducing gross disproportionality at the final stage does not do enough to address the dilemma posed by factual uncertainty that precaution is supposed to solve.
A third recommendation is to assume a factual premise in cases of evidentiary uncertainty and predicate proportionality analysis on that assumption. A court might assess the government’s justification for a law and modulate how much deference it affords to the legislature under the assumption of either a ‘worst-case’ scenario that the harm the government seeks to prevent will be grave or a ‘best-case’ scenario that the harm will be minimal.Footnote 62 Consider the law in Harper limiting third-party election ads. In the worst-case scenario, wealthy private interests would hijack elections. In the best-case scenario, the ads would have no meaningful impact on elections.
Now, these assumptions must be carefully calibrated. The worst- (or best-case) scenario that a law might presumptively address cannot simply be one where the targeted harm is extremely serious (or minimal) without regard to the harm’s likelihood.Footnote 63 Otherwise, a law targeting very serious yet highly speculative harm might easily be justified, no matter how severe its impact on constitutional interests. In the same vein, a law targeting a minimal harm might never be justified despite the harm’s immanence and the public interest in averting it. Conversely, each assumed scenario cannot only specify the likelihood of a harm without considering its severity.Footnote 64 The public interest in preventing a catastrophically large harm may be capable of justifying limits on rights even if the harm is unlikely to occur.Footnote 65 Or the value of protecting rights may be greater than the importance of preventing minor, yet highly probable harm. Worst- and best-case assumptions must therefore account for both the gravity and likelihood of harm; it is a mistake to exclude either variable.
However, the difficulty implicit in the foregoing is that if a court merely assumes the worst- or best-case scenario, it invites a kind of ‘circular’ or ‘tautologous’ reasoning into the proportionality analysis.Footnote 66 If the worst-case scenario is assumed, including both the severity and the certainty of the harm to be prevented, limits on rights imposed by a law will invariably tend to be justified. If the best-case scenario is assumed, including both the insignificance and remoteness of the harm, the limits will tend to be unjustified. Depending on the premise put in place, therefore, the outcome is self-fulfilling, and the premise itself rather than proportionality reasoning does all the work.Footnote 67
Let us summarize the takeaways that have emerged from the discussion up to now. To defensibly incorporate precaution into proportionality, we need a modest, minimally revisionary adjustment to the Oakes test that explains how each stage of the test operates in factually uncertain contexts and that requires courts to evaluate both the severity and likelihood of the harm a law targets without adopting preordaining assumptions. The evaluation must respond to evidence placed before the court that amounts to more than a bald assertion by the government of serious and certain harm while not necessarily satisfying the strict standard of clear and cogent evidence. Our approach aims to take all these lessons seriously.
The risk-based approach
Under risk assessment, as we construe it, the level of deference courts must show to the government when applying proportionality scales according to the gravity and likelihood of the harm that a law aims to mitigate. On the least deferential end of the scale, gravity and risk are both low, while on the most deferential end, they both are high. Middling levels of deference are warranted where one or the other variable is high. For example, the harm of spreading COVID-19 is extremely severe; while the risk of infection is lower outdoors, if the limited available evidence shows the risk is non-negligible, the state’s choice to restrict both indoor and outdoor gatherings may very well be entitled to deference.
On this view, there is a mutually reciprocal relationship between evidentiary standards and judicial deference. The state must provide some, albeit limited, evidence demonstrating the severity and likelihood of harm. Insofar as it does, it is absolved of the need to establish these facts with total credence, and it need not meet the legalistic empirical standard of clarity and cogency. The approach we are recommending thus treats evaluations of seriousness and risk of harm as legal principles that not only add structure, rationality, and clarity to judicial decision-making, but also political resources that create opportunities to empower legislative decision-making and enhance the freedom of policy choices that would otherwise be obstructed by the absence of factual certainty.Footnote 68
Despite its professed novelty relative to other scholarly proposals, risk management is not unprecedented and can, to some extent, be extricated from current Canadian constitutional jurisprudence. For example, when considering whether there are alternatives to fully prohibiting hate speech that limit free expression less drastically, the Supreme Court of Canada noted in the Whatcott case that the government only needs to show that it has a reasoned apprehension of the harm caused by hate speech – namely, the delegitimization and vilification of vulnerable social groups – and that the prohibition will mitigate that harm. In the Court’s view, the gravity of the potential harm of hate speech makes it necessary to defer to the government’s apprehension of the risk of that harm and not insist on definitive proof of the causal link between hate speech and the harm in question.Footnote 69 Gravity and risk are, thus, the two variables that drive the assessment of less drastic alternatives.
It is nonetheless illuminating to grasp how risk analysis differs from similar views already found in comparative constitutional theory. It may be likened to Aharon Barak’s concept of ‘principled balancing’ under the final step of proportionality. This concept seeks to give more concrete and determinate content to the abstract injunction to balance a law’s overall positive and negative effects while still being general enough to provide non-ad hoc guidance across distinct fact patterns. Barak writes that under the final step, the deleterious effects of a legal limit on constitutional rights must be balanced against ‘the marginal social importance of the purpose as interpreted against the background of its content and the degree of urgency in its realization, the probability of its realization and the harm that would be incurred should the purpose not be realized, and the probability of the occurrence of such damage’.Footnote 70 As we explain below, risk analysis differs from principled balancing in that it is not restricted to proportionality strictu sensu but is activated at each stage of the proportionality test. This is also what differentiates it from the Supreme Court of Canada’s discussion of rationalist standards under Oakes in Whatcott, which was restricted to minimal impairment.
Risk assessment is also analogous to what Robert Alexy calls the ‘epistemic law of balancing’.Footnote 71 Alexy acknowledges that in factually uncertain contexts, the legislature is entitled to exercise ‘empirical epistemic discretion’.Footnote 72 The state need not show with absolute certainty that the objectives of laws that limit constitutional rights will be achieved. However, it must also do more than dubiously assert that they will be. The solution for Alexy is for judges to mandate a level of proof that corresponds to the extent of the law’s interference with constitutionally protected interests. The epistemic law of balancing thus holds that ‘[t]he more heavily an interference in a constitutional right weighs, the greater must be the certainty of its underlying premises’.Footnote 73 This solution aims to strike a balance between the ‘substantive principle’ of respect for constitutional rights and the ‘formal principle’ of democratic legitimacy in legislative decision-making.Footnote 74
Alexy’s view is subject to detailed debate and refinement that is beyond the scope of the present study.Footnote 75 For our purposes, it is important to note that risk assessment agrees with Alexy’s epistemic law of balancing in affording some latitude to the legislature’s empirical epistemic discretion. However, because it focuses proportionality reasoning in part on the foreseen risk of a prospective harm that has some anticipated magnitude, it provides guidance for courts and legislatures even in cases where a law is enacted prophylactically to prevent harm that has not yet concretely materialized. For example, risk analysis is available to courts when conducting ‘abstract review’Footnote 76 if questions about the constitutionality of an unenacted law are referred to the judiciary. It is available to empower legislatures when they perform ‘pre-enactment political rights review’Footnote 77 on a proposed bill during the drafting process. However, Alexy’s law of balancing is unavailable in these circumstances since it requires some definite knowledge, or at least an accurate prediction, of the seriousness of an interference with a constitutional right that has previously come to fruition. This is because it tightly ties judicial deference to the weight of an established interference. Therefore, it is less broadly applicable than risk management as a guideline for controlling proportionality reasoning.
Others have also argued that, even where Alexy’s epistemic law of balancing applies because an antecedent interference with a right is quantifiable, it will be impossible to justify the interfering measure if the magnitude and likelihood of the harm it targets is uncertain, but the interference is very serious. Tying proportionality tightly to the weight of an interference stacks the deck in favor of the substantive principle of respect for rights in the contexts of uncertainty.Footnote 78 However, risk analysis conversely makes room for the formal principle of democratic legitimacy by empowering legislative decision-making to sacrifice rights, even extremely seriously, in the face of uncertainty in appropriate circumstances.
Setting aside comparative constitutional theory, something like risk management is discernable in tort law doctrine across jurisdictions. It resembles tools used for defining the standard of care and determining whether a defendant has acted unreasonably or negligently in harming the plaintiff. Echoing the well-known American ‘Hand Formula’, the Supreme Court of Canada has stated that the standard of reasonable care depends on ‘the likelihood of a known or foreseeable harm, the gravity of that harm and the burden or cost’.Footnote 79 That is, under negligence law, a reasonable person is one who limits their own liberty by taking such precautions designed to avoid incidentally harming others that are proportionate to the gravity and likelihood of that harm.Footnote 80
In private adjudication, judges have found themselves adept at evaluating the gravity and risk of harm in the absence of definitive proof of either. For example, the British Columbia Court of Appeal has held that the risk of injury to pedestrians when a plainly visible 45-foot pole was left on a sidewalk was ‘small’ when an unobstructed passage around the pole existed. The defendant utility company was therefore not negligent in leaving the pole unattended when the plaintiff tripped over it.Footnote 81 The same court reached a different conclusion where a defendant poured motor oil over an open fire, finding that such conduct posed a ‘significant’ risk of ‘severe’ injury.Footnote 82 Just as courts have been prepared to ascribe qualitative markers to the significance of risk and severity of harm, they have been capable of balancing these variables to achieve an overall assessment of proportionate precautions needed to avert the harm at issue. Thus, even a small risk can result in liability if precautions are not taken when the potential loss is substantial. Conversely, where the risk is minor, the foregoing precautions may be acceptable even if the possibility of serious harm exists.Footnote 83
In the same vein, we maintain that whether governments may justifiably enact precautionary measures that limit constitutional freedoms depends on the gravity and likelihood of the harm the government aims to prevent. In this sense, situating risk analysis inside Oakes should not be seen as revisionary. It is nothing more than taking a familiar analytical lens from the adjudication of private wrongs and looking through it to guide constitutional adjudication in the same way that, for example, the analytical lens of tort law has been applied to shape the so-called ‘constitutional torts’.Footnote 84
Risk management principles also surface in Canadian criminal law. First, substantive issues in the criminalization of risky activities should be considered. For instance, the Supreme Court of Canada held in R v Mabior that people with sexually transmitted infections can be convicted of aggravated sexual assault where there is a ‘significant risk’ of transmitting the disease to their sexual partner and they fail to disclose their infection.Footnote 85 The meaning of ‘significant risk’ depends on both the risk of contracting the infection and the seriousness of the infection if contracted.Footnote 86 The Court in Mabior cited various studies that put the risk of contracting HIV anywhere between 0.05% and 0.26% after unprotected vaginal intercourse. This was sufficient to constitute a significant risk of bodily harm, notwithstanding the relatively modest risk of transmission. Contrast this finding with a lower court decision, which the Court endorsed, deciding that a considerably higher 1%–2.5% chance of contracting Hepatitis C was ‘so low that it cannot be described as significant’.Footnote 87 As the Ontario Court of Appeal recently noted, ‘[t]he more serious the harm, the lower the probability of transmission needs to be’.Footnote 88
Second, many rules governing the criminal process require courts to determine whether it is proportionate to deprive an accused of liberty out of precaution to protect public safety.Footnote 89 In the bail context, detainees may be denied bail where there is ‘substantial likelihood’ they will re-offend.Footnote 90 The meaning of ‘substantial likelihood’ is not tied to the likelihood of offending alone:
Rather, it must be weighed against the gravity of the harm that will ensue if the risk comes to pass. For example, even a very grave risk that an incorrigible petty thief will shoplift again if granted bail is one that the court might be willing to take when balanced against the accused’s constitutional right to reasonable bail. On the other hand, where the anticipated harm is very grave, a more remote risk may be sufficient to meet the test of substantial likelihood.Footnote 91
Similar considerations are raised in the designating and sentencing of dangerous offenders. Where an offender exhibits a pattern of violent conduct, a court may impose an indeterminate prison sentence if it is satisfied, beyond a reasonable doubt, that there is a ‘likelihood’ he will cause ‘death or injury’ or inflict ‘severe psychological damage on other persons’ when released.Footnote 92 Thus, the court must assess the likelihood and seriousness of future offending when deciding whether someone should be imprisoned indefinitely despite the impossibility of predicting with certainty whether the person will reoffend.Footnote 93 Given that risk management principles pervade the criminal justice system, it stands to reason that the same principles will prove useful when adjudicating constitutional disputes.
How to incorporate risk analysis into Oakes
We have argued that risk analysis is juridically familiar and pays heed to both the severity and the probability of the harm the government seeks to prevent. We now explain how it should operate within the Oakes test. Rather than substantially altering the analytical steps under Oakes in the way that alternative proposals do, risk assessment should be anchored in each stage of the standard proportionality inquiry. By providing added structure, order, and determinate content to each step of Oakes, it does a better job at containing result-driven judicial deference (or defiance) than simple reliance on a priori standards of logic and common sense.
Under Oakes, the state must first establish that the goal of a rights-limiting measure has sufficient importance. Courts must identify the goal with specificity; if it is too broadly cast, its value may be exaggerated, and later stages of the proportionality analysis may be compromised.Footnote 94 This directive also mandates the precise specification of the goal’s normative importance. When there is factual uncertainty about the harm the government aims to mitigate, risk analysis assists in precisely defining the normative importance of the objective of mitigating it and controlling the amount of curial deference the state is entitled to in this regard. It does so by concentrating the court’s attention on the nature of the harm and the likelihood it will occur. To illustrate, preventing a grave harm will be more important than preventing a small one, assuming the risks of each are the same. Preventing a small harm that has virtually no chance of materializing is hard to characterize as valuable. However, preventing a catastrophic harm that is unlikely to occur, or a smaller but more palpable and immanent harm, may be sufficiently important.
The limit on rights must also be rationally connected to the aim of the impugned law. This stage of Oakes is sensitive to evidentiary uncertainty. All that is required is some satisfactory proof of a causal connection between limit and aim (what the Supreme Court of Canada refers to as ‘reasoned demonstration’). The state need not prove that the law will certainly advance the aim, only that it is reasonable to suppose it could.Footnote 95 Here, the proposed model plays a role in ascertaining the requisite threshold of satisfactory proof that rises above a unilateral declaration by the state that the causal connection exists. The threshold correlates with the gravity and likelihood of the harm targeted by the state. It is less stringent where both gravity and likelihood are high, more stringent where these variables are low and middling when only one variable is high.
The next stage is minimal impairment – are there alternative means that interfere with rights less while still attaining the government’s objective?Footnote 96 In the contexts of uncertainty, risk assessment operates at this stage by regulating how much deference the state is entitled to when a court considers whether a proposed alternative is just as effective as the state’s chosen measure. The greater the severity and/or probability of that harm, the more deference is warranted to the state’s assertion that its chosen means is necessary, and that alternative means are ineffective.
The final stage concerns whether the overall benefits of realizing the end of the impugned law outweigh the overall costs of the rights limitation.Footnote 97 This stage requires courts to make ‘difficult value judgments’.Footnote 98 However, even here normative evaluations are affected by factual uncertainty. The subjectivity of overall value judgments can be minimized by construing the benefits of a law in terms of the seriousness and risk of the harm it seeks to alleviate. As suggested by Barak’s concept of principled balancing, if gravity and/or risk are high, the easier it will be to conclude that the law has highly salutary effects compared to its deleterious effects. By contrast, the salutary effects will be seen as lesser if gravity and/or risk are low. That said, when gravity and/or risk are low, a law without significant salutary effects may still be proportionate if its deleterious effects are negligible, whereas as Alexy’s epistemic law of balancing suggests, if its deleterious effects are significant a higher degree of gravity and/or risk may be warranted. In all these ways, risk assessment helps to articulate guidelines that manifest throughout the entirety of the Oakes test. It renders more precise the state’s permission to limit rights when promoting the public good out of precaution.
We acknowledge that risk analysis does not fully resolve the problem of indeterminacy in proportionality analysis. Eliminating indeterminacy may well be an impossible ideal. However, reducing or minimizing it is certainly preferable, not only to the highly revisionary approaches found in the comparative constitutional literature, but also to the unstructured approach that currently prevails in Canada. We do not defend risk management as a panacea for judicial subjectivity. We claim that it brings about a net decrease in vagueness as compared to the use of rationalistic standards such as a reasoned apprehension of harm. In addition, there is a good reason to think that there would be a net decrease. As we saw, courts have been able to characterize the gravity and the probability of harm in a suitably objective fashion in civil and criminal law. Similarly, in constitutional law, it seems possible for courts to evaluate these magnitudes using a serviceable scale of ‘light, moderate and serious’, as proposed by Alexy.Footnote 99 For instance, the gravity of the harm to public health caused by the transmission of COVID-19 could be characterized as severe even if scientific projections about the virus’s mortality rates are elusive. By comparison, to take an example from an Ontario Court of Appeal decision, the harm that pit bulls pose to society seems light to moderate.Footnote 100 Analogous claims can be made about the degrees of risk and probabilities. Furthermore, greater credence about one variable on the gravity/risk ledger may compensate for less credence on the other. If a potential harm is obviously severe, ambiguity as to its potentiality may be tolerable. Or, if the risk is high, it may not be so important to ascertain with total accuracy just how severe the potential harm is.
In sum, inconclusive evidence may inevitably invite challenges of vagueness and subjectivity into proportionality reasoning for precautionary laws. However, this does not preclude investigations of the relative merits of distinct tools designed to meet these challenges. Our position is that risk assessment is preferable in this respect to the extant tools that have emerged.
Conclusion
Government policy is rife with unknowns – unknown harms and unknown answers. Just as often, measures designed to mitigate these unknowns implicate citizens’ fundamental rights. When considering diverse and unpredictable areas such as public health, child protection, electoral integrity, crime reduction, national security and traffic safety, when can the government limit rights to advance collective values without the benefit of evidentiary certainty? It may appear that the state can either never justify such limits or, if it has an open-ended permission to limit rights out of precaution, it will be too easy to do so. Worse still, if the lack of knowledge means that courts must defer to the legislative permission, courts may defer selectively when doing so is congenial to judges’ subjective preferences, and there are otherwise no transparent strictures on the exercise of that permission.
In the past, the Supreme Court of Canada has sought to escape the dilemma of factual indeterminacy by accepting relaxed evidentiary standards when applying the Oakes test under Section 1 of the Charter, but its judgments attracted criticisms of ad hoc discretion about when to defer. The lack of content for filling in the contours of the legislature’s precautionary permission has allowed this criticism to be rekindled following the COVID-19 litigation in Canada. We have advanced a risk-based approach – which sees the appropriate level of deference as corresponding to the seriousness and probability of harm that the government addresses – to provide that content at every stage of Oakes and thereby minimize problematic judicial discretion. This approach is an incremental adjustment to proportionality in accordance with the familiar legal principles found in Canadian tort law and criminal law. Given its promise for navigating out of the dilemma posed by factual uncertainty, we commend it as informing the interaction between proportionality and precaution beyond the Canadian horizon.
Competing interest
The authors declare none.
Open access
