Sean Rehaag on Revisiting the Luck of the Draw

Back in 2012 I posted about a paper by Sean Rehaag which identified “wild variations in the treatment of leave and judicial review applications by different Federal Court judges”. Reforms to the leave and judicial review process followed.

Professor Rehaag has now returned with a new study, “Revisiting the Luck of the Draw“, covering an even larger dataset and comes — despite the reforms — to similar conclusions:

The data considered in this article shows that little has changed since the earlier luck of the draw study.

Notwithstanding the Court’s attempts to increase convergence in leave grant rates, from 2013 to 2016 a claimant’s chances of success in securing leave and ultimately succeeding with their application for judicial review continued to depend a great deal on who was assigned as the leave judge.
The ranges in grant rates across leave judges are substantial, with leave grant rates going from 5.3% to 49.2%, and with overall success rates going from 1.8% to 22.8%. In other words, from 2013 to 2016, if a claimant was lucky with leave judge assignment, they could be up to 11.1 times more likely to succeed with their application than if they were unlucky with leave judge assignment. And, perhaps most importantly, this is not a phenomenon restricted to a handful of outlier judges. Rather, leave grant rates vary across the board.
The same is true at the JR stage in cases where leave is granted. This study has shown that from 2013 to 2016 there continued to be a remarkable range in grant rates on the merits depending on who served as the JR judge, from 13.8% on the low end, to 95.1% on the high end – and even if Justice Campbell is set aside as an outlier on the high end, the range still runs from 13.8% to 69.8%. In short, this study has found that refugee claimants whose applications for judicial review are denied continue to have good reason to wonder whether this was because of the facts of their case and the law, or whether they simply lost the luck of the draw.

Professor Rehaag recommends that arbitrariness could be reduced by assigning the same judge for the merits and leave stages, a measure that would not eliminate arbitrariness (given the differences in grant rates at the merits stage) but would certainly reduce it by removing the most objectionable discrepancies from the existing system:

Right now, the leave requirement acts as a gatekeeper to having the matter decided on the merits before the Federal Court. As this study and others have shown, that gatekeeping function is being applied in an arbitrary manner. Judges who are reluctant to grant leave are denying leave to cases that might otherwise have a good chance of succeeding.
If, by contrast, the leave judge and the merits judge were the same judge, the leave requirement would instead act as a tool for effectively allocating judicial resources. If a judge is of the view that a case has no reasonably arguable case and the same judge would ultimately hear that case, then there would be little value in holding a hearing, as the case would be destined to fail before that judge. This would not, of course, eliminate the luck of the draw. Outcomes would continue to differ depending on which judge was assigned. It would also not eliminate some of the other problems with the leave requirement – including, for example, the lack of transparency caused by judges not giving reasons when they deny leave.

Another recommendation is that judges should take a different approach to determining whether to grant leave:

A preferable approach, in my view, would be for the leave judge to consider not whether they think a reasonably arguable case has been made, but rather whether any of their colleagues – including those who are most permissive in granting leave – might be of the view that the applicant has made out a reasonably arguable case. Judges who adopt this approach would only deny leave in cases where that judge thinks that all other judges on the Court would agree that there is no reasonably arguable case. If the leave judge properly applies this approach, then leave will only be denied where there is no purpose to proceeding to a hearing because the case would eventually be denied on the merits by whichever judge was assigned at the JR stage.

This looks like a Bayesian approach, with judges required to adjust their priors in view of the views they would attribute to their colleagues, something which would be facilitated by regular reading of others’ decisions (and by regular interactions with colleagues).

The most arresting of the implications Professor Rehaag draws from his analysis is that the leave regime might be vulnerable to a constitutional challenge:

It may well be, however,that constitutional litigation will force the matter. This study, and others like it, showthat the leave requirement has proven to be applied unfairly over a long period of time, despite attempts by the Court to improve consistency. Section 7 of the Canadian Charter of Rights and Freedoms provides refugee claimants the right not to be deprived of life, liberty and security of the person, other than through processes that comply with the principles of fundamental justice. By imposing an arbitrary barrier on access to judicial oversight in refugee adjudication, the leave requirement arguably violates that right, and may therefore be vulnerable to a constitutional challenge.

Given the evolution of the s. 7 jurisprudence, this possibility should not be dismissed out of hand.

Finally, the more tech savvy amongst you will enjoy Professor Rehaag’s description of the code he used to scour the Federal Court’s dockets:

While the code was highly accurate, it should be noted that errors were not random. The code struggled with two types of problems. The first type of problem related to complex cases, such as cases with multiple procedural motions or where steps in the process were overturned (e.g. leave is denied, but there is a motion to reopen the leave determination). My research assistants were better at coding these cases because they were able to first understand what was going on in the case and then determine which parts of the docket contained the datapoints of interest – and even where they were unable to do this, they were at least able to flag the case for my review. The code, by contrast, treats such cases as usual, and just looks for the regular patterns, sometimes confusing one of several steps in the process for the leave or judicial review outcome. The second type of problem related to information being entered in the docket in a novel way – whether due to typographic errors or just a typical ways of recording standard information. Human research assistants had no trouble coding such cases because they were able to work out from the context what the person entering information into the docket had intended. The code, however, lacking this contextual understanding, was not able to accurately gather data for such cases.
Over to the Federal Court!


This content has been updated on September 21, 2018 at 15:44.