Police discipline in British Columbia · the published record, 2010–2025
1,243 allegations · 727 cases · 15 municipal agencies · 576 adjudication documents · 2010–2025
Where the data comes from, and how the coding was checked.
The source is the Discipline Decisions Digest of the Office of the Police Complaint Commissioner of British Columbia, the civilian body that oversees the province's municipal police. The digest summarises every concluded case. This site covers the 1,243 allegations concluded between September 2010 and March 2025, across 727 cases and 15 agencies. The unaltered export, and the two reference documents behind it (the OPCC's bulletin on internal discipline, and the province's Police Resources report with each department's officer count), are in the archive on the data page. The OPCC also publishes the underlying case files on its hearing and review pages, 576 documents in all, decisions along with notices, submissions and rulings; the archive lists a URL for each.
The OPCC oversees municipal police only. Most policing in British Columbia is done by the RCMP, which answers to a separate federal process, so no RCMP officer appears anywhere on this site.
We added two layers to what the OPCC published. First, a classifier read each published summary and assigned one of fourteen conduct categories. Second, a flag marks gender-based violence under four stated rules, with the supporting passage recorded for every allegation it marks. Everything else, the headings, sub-labels, measures and case numbers, is the OPCC's own.
The conduct categories and the flag were assigned by a machine reading the published summaries. Nobody has coded all 1,243 allegations by hand, so there is no answer key to compare against. Each of the four checks below tests one way the coding could be wrong, and states what passing does and does not establish.
Every label carries a quote the classifier says it copied from the source. The quote appears word for word in the source for 1186/1243 (95.4%) of allegations, and appears once punctuation differences are set aside for 1204/1243 (96.9%). A quote that is not in the source was invented, and the label under it cannot be trusted. This check catches invention; it does not show that a real quote supports the right category.
The OPCC attaches a keyword of its own to each allegation. For 47 of those keywords there is only one category ours could reasonably map them to, so the OPCC's choice can serve as the answer. 609 allegations carry one of those 47 keywords. The classifier gave the expected category to 566/609 (92.9%) of them, and made it the main category for 549/609 (90.1%).
The first reading flagged 110 allegations as gender-based violence. A second, independent reading then went back through those 110 with the opposite instruction: reject the flag wherever the text allows it, and reject when uncertain. It upheld 90/110 (81.8%); those 90 are rule 1 on the flag's page. A check built to reject can only shrink the set. It weeds out weak flags, and it cannot find gender-based violence the first reading missed, so the 90 is a floor rather than a full count.
150 allegations were re-classified from scratch, without sight of the first label. The two readings agreed on the primary category for 137/150 (91.3%) (Cohen's kappa 0.90). Agreement shows the categories are drawn clearly enough to apply consistently.
The digest publishes concluded cases with at least one proven allegation. Complaints dismissed earlier are not in our database. The published summary can leave things out: file 2015-11048-03 says "unwanted physical contact" where the OPCC's own sub-label reads sexualized harassment. Officers are named where the OPCC named them and are otherwise "the member"; we did not try to match officers across files, so an officer who appears in two files counts twice. The suspension-day and duration columns cover only the allegations where the digest records them.