Methodology: Federal Court Refusal Statistics
The corpus
The statistics are computed from a research collection of 3,967 Federal Court of Canada decisions spanning 1992–2026, assembled for the analysis engine behind our tools. Decisions are public documents; we publish only aggregate statistics derived from them, not decision text.
Classification
Each decision was labeled in two independent automated passes (large language models with structured output), the second using stricter criteria designed to exclude look-alike non-immigration cases such as employment-insurance, tax, and public-service staffing disputes. A decision counts as an immigration case only when both passes agree. Labels cover the area of law, the main disputed ground, the judicial-review outcome, and whether occupation classification or employment evidence mattered. Samples were manually checked against the underlying decisions during development, and identified error patterns fed back into the second pass.
Limitations — read before citing
These are statistics about litigated cases, not about all refusals: only a small, self-selected fraction of refused applicants seek judicial review, so success rates in court say nothing about the error rate of refusals overall. "Disputed ground" reflects the main issue argued, not every issue in the file. Older decisions are overrepresented relative to recent years. Automated classification retains a residual error rate despite the two-pass design and manual checks. None of this is legal advice.
Citing these statistics
You are welcome to cite these figures with attribution: "ScoreRoute Canada, analysis of 1,803 Federal Court immigration decisions (1992–2026)", linking to this page. The dataset behind the aggregates is not redistributed.
Educational statistics derived from public Federal Court decisions. Not legal advice, not a prediction of any individual outcome, and not affiliated with IRCC or the Courts.