Home / Corrections
Corrections
This site publishes specific figures about named law enforcement agencies. If one of them is wrong, we want to know — and when we fix it, we say so here rather than quietly editing the page.
Reporting an error
Email corrections@policecensus.com. The fastest thing you can include is the agency name and its ORI — the FBI identifier printed at the bottom of every agency page — and the figure you believe is wrong.
If a number here does not match an agency’s own records, the discrepancy is usually in what was filed with the FBI rather than in this site’s arithmetic — Police Census reproduces what the agency reported and does not adjust it. That is still worth telling us: where the filed figure is the problem, we would rather say so on the page than leave a number standing that the agency itself disputes.
What we correct, and how
A correction is published below with what the page said before and what it says now. The alternative — editing the page and saying nothing — leaves anyone who already read the old version still believing it.
Not every disagreement is an error. Where a figure is correctly reproduced from its source and the dispute is with the source, we say that on the page rather than changing the number. Where this site got it wrong, it goes in this log.
Published corrections
| Date | Where | What it said | What it says now |
|---|---|---|---|
| 2026-10-11 | /deaths, /sources, /about and the deaths block of every agency page that shows one found internally | Deaths during police encounters were over-counted for about six weeks. The site holds two sources, Mapping Police Violence (2013–2026) and Fatal Encounters (2000–2021), which describe many of the same deaths in 2013–2021, and the query layer never excluded the duplicates. /deaths said 47,047 deaths; it now says 37,956. The Los Angeles Police Department page said 716; it now says 569. The Phoenix Police Department page said 434; it now says 311. In all, 2,558 agency pages were inflated. The per-capita ranking on /deaths was not merely inflated but in the wrong order, because the duplicates fell unevenly across agencies. Separately, every deaths figure was captioned as coming from Mapping Police Violence alone, and /sources and /about listed it as the only source. That credited one organisation with records it did not produce: 1,919 agencies have no Mapping Police Violence records at all, and Fatal Encounters, which supplies about two thirds of the records, was named nowhere. | The total is 37,956 (15,550 + 31,497 − 9,091): 9,091 Fatal Encounters records are excluded as duplicates of Mapping Police Violence records, matched on state, exact incident date and the first and last name of the person. The excluded records are kept and flagged, not deleted, and a death recorded with a different date or spelling in the two sources would not be matched. The ranking on /deaths is recomputed on the deduplicated counts. /deaths, /sources and /about now name both sources with their real date ranges, and /sources gives Fatal Encounters’ stated terms (“we let anyone use the data for any reason for free”). Each agency page now says its figure draws on one or both of the two sources and does not attribute it to either alone. The same pages previously said Mapping Police Violence data was used with permission; they now say it is attributed to its compiler and that the written scope of the site’s right to display it is being confirmed. Raised internally, not reported by a reader or an agency. |
| 2026-10-07 | /oversight, /agencies/ma/worcester-police-department and /agencies/nj/newark-police-department the source changed | The Worcester Police Department page carried a Federal oversight block labelled “Investigation”, and the Newark Police Department page one labelled “Enforcement” — stating as a current fact that each department was subject to an open Department of Justice civil-rights matter. | Both are labelled “Closed”, which is the word the Justice Department’s own Special Litigation Section register now uses for each matter. The rows are kept rather than deleted: removing them would imply these departments were never subject to federal oversight, which is false in the opposite direction. The label is reproduced verbatim and is not translated into a characterisation such as “decree terminated”. Found by the nightly gate that re-fetches the DOJ register and compares every stored label against it; it first failed on 2026-10-05, so the stale labels were live for about two days. The gate reports a label change distinctly from a missing row precisely because a closed matter and an absent matter are different facts. |
| 2026-08-30 | /law/cases/utah-v-strieff and /law/cases/barnes-v-felix found internally | Both pages gave a flat citation count — "Cited by 425 later opinions" for Utah v. Strieff and 122 for Barnes v. Felix — which reads as the number of times the case has been cited. | Both figures are shown as lower bounds, "at least N", with the reason stated on the page. CourtListener holds two separate records for each of these cases — same caption, same filing date — and the opinions citing them are split between the two. Strieff’s other record carried 373 and Barnes’s carried 1. The counts are not added together, because the two sets may share citing opinions and the sum would be a total nobody measured. The number shown is the one on the record this site adjudicated, which is true and incomplete, so the page now says so. Found while building the automated citation refresh, which asks CourtListener for each citation and refuses any answer the source itself marks as having multiple matches. |
| 2026-08-30 | /oversight found internally | The page said the Puerto Rico Police Department and the Virgin Islands Police Department had no agency page because "Puerto Rico and the Virgin Islands are not in the FBI agency register this site is built on, which covers the 50 states and DC only". | The FBI register does cover them. Asking the CDE agency API for territory codes returns Puerto Rico (PRPPR0000), both US Virgin Islands districts (VI0010000 St. Thomas, VI0010300 St. Croix) and Guam (GM0010000). They were missing because the ingest here only ever requested 51 state codes, not because the FBI omits them. The register now holds 19,625 agencies including the territories, Puerto Rico has an agency page carrying its DOJ matter, and the Virgin Islands row is left unattached for a different and stated reason: the DOJ case covers one territory-wide force while the register splits it into two district agencies. |
| 2026-08-26 | /law — the 24 pages for 28 C.F.R. Parts 20 and 23 found internally | Every section of 28 C.F.R. Part 20 and Part 23 was published under the heading "Crime reporting to the FBI", which described them as the regulations behind the UCR and NIBRS crime statistics this site publishes. | Part 20 is headed "Criminal history records" and Part 23 "Criminal intelligence systems", which is what each part says it governs — rap sheets and the Interstate Identification Index in the first case, intelligence databases in the second. Neither is the UCR/NIBRS regulation. There is in fact no CFR part implementing UCR/NIBRS: 28 U.S.C. § 534 is the authority and participation is FBI policy rather than regulation, which the /law page now states rather than filling the gap with an adjacent part. |
Source
How each dataset is collected, what it does and does not claim, and when it was last refreshed is on Method & sources.