Bias and Statistical Evidence: ALA15 v Minister for Immigration and Border Protection [2016] FCAFC 30

There is a new decision from the Full Federal Court of Australia on whether and how statistical evidence can be used to demonstrate bias (in particular, prejudgement of relevant issues): ALA15 v Minister for Immigration and Border Protection [2016] FCAFC 30. Readers will recall that this question has also been raised in Canada (unsuccessfully) and Ireland. Here, the applicant was unsuccessful in seeking the recusal of a judge of the Federal Circuit Court.

The grounds, contained in the affidavit of a court reporter, are laid out at paragraph 11 of the Court’s decision:

(a)    He identified 286 decisions of Judge Street during the relevant period, of which 254, or 88.81% were in the area of immigration law where the Minister for Immigration and Border Protection was the respondent (immigration judgments).

(b)    In all 254 or 100% of the immigration judgments, they were, or appeared to be, delivered ex tempore.

(c)    Only in two of the 254 immigration judgments, or 0.79% of the immigration judgments, Judge Street found in favour of the applicant against the respondent Minister for Immigration and Border Protection.

(d)    In 252 out of the 254 immigration judgments, or 99.21%, Judge Street found in favour of the respondent Minister for Immigration and Border Protection.

(e)    There were only two judgments where the primary judge found in favour of the applicant. In ABT15v Minister for Immigration and Border Protection [2015] FCCA 1051, the Minister for Immigration and Border Protection conceded that there was an error and in Kautoga v Minister for Immigration and Border Protection (No 2) [2015] FCCA 1679, that there was jurisdictional error.

(f)    In at least 163 of the 254 immigration judgments, or 64.96%, the immigration judgments were given at the first court date. In approximately another dozen cases it was not possible to tell if they were delivered at the first court date or not.

(g)    During the relevant period, the remaining eight judges of the FCCA in Sydney delivered 309 immigration judgments, or 54.89% of the immigration judgments in the Sydney Registry of the FCCA, whilst the primary judge personally delivered 45.11% of the immigration judgments.

(h)    The most recent Annual Report of the Migration Review Tribunal (MRT) – Refugee Review Tribunal (RRT) disclosed that 10.8% of MRT decisions and 12.2% of RRT decisions were set aside, compared with only 0.79% being set aside by the primary judge on judicial review.

However, this statistical evidence was insufficient to demonstrate bias, especially because the well-informed observer (whose perspective a court has to adopt when bias is alleged) would have to consider not only raw statistics but also the more general context of the underlying cases:

[F]or such raw statistical material to be attributed to the hypothetical observer, it normally would need to be accompanied by a relevant analysis of the individual immigration judgments determined by the primary judge in order that the statistics were placed in a proper context. Absent such analysis, the hypothetical observer would not be able to make an informed assessment of the significance of the raw statistics. It may be, for example, that a close analysis of some, many, or all of the relevant judgments reveal that they had been determined on a reasonable and plausible basis. And, even if some or all of the judgments were wrongly decided, that may be the consequence of human frailty on the part of the judge and not prejudgment, a consideration which a fair-minded lay observer would take into account (at para. 38).

On its own, the fact that one judge might be more likely than another to find in favour of or against an applicant is not evidence of bias (at para. 39).

Ultimately, “the mere fact that a particular judge has decided a number of cases, the facts and circumstances of which are unknown, one way rather than another, does not go any way to assisting the hypothetical observer making an informed assessment as to whether that judge might not bring an impartial and unprejudiced mind to the resolution of the question in a particular proceeding before that judge” (at para. 41).

The applicant did try to tie the statistical evidence to particularly strong language found in the judge’s opinions. But this too was insufficient to demonstrate bias:

Insofar as the applicant’s case relies on the assertion that the primary judge had repeatedly used phrases such as “no substance” or “it is clear” in many of his immigration judgments, no evidence was adduced to substantiate this claim, which was only raised in the applicant’s written submissions. In any event, and perhaps more significantly, even if it be the case that the primary judge has repeatedly used those expressions, that is not probative of whether the hypothetical observer might apprehend that the primary judge might not bring an open mind to bear in determining the applicant’s proceedings. Such phrases are not uncommonly used by judicial officers. Their use, even frequently, would not of itself indicate to the fair-minded and informed hypothetical observer that a judge might be biased, at least without the observer knowing more about the context in which the phrases were used (at para. 46).

The unsurprising conclusion, then, is that statistical evidence alone is not enough to demonstrate prejudgement of relevant issues by decision-makers who, after all, are presumed to be impartial. A deeper question is whether the threshold should be as high as the Australian and Canadian courts have set it. Where decision-makers demonstrate marked tendencies of decision one way or another, the risk of similarly situated individuals being treated differently is quite high, with all the rule-of-law concerns such inconsistency provokes. And where such decision-makers are independent, in the sense of being protected from political interference, it is difficult to see how the political branches might respond to the problem. If so, judicial oversight ought arguably to be more exacting.

This content has been updated on March 10, 2016 at 03:29.