Suppression and privacy

What is published

This site publishes aggregate counts only: how many encounters fall into each category of a variable, sometimes split by year. Dates are reduced to the year. Every chart states the number of records it covers and the number excluded.

What is never published

The Lab does not publish names of any person (subjects or officers), case identifiers from external databases, exact dates, precise locations, agency names, case narratives or source documents, or any field that could directly identify an individual or case from public sources.

Restricted case-level data

A de-identified case-level dataset for each snapshot is deposited with Zenodo under restricted access. The deposit does not include names, external database identifiers, exact dates, sub-county geography, or case narratives. It does include the Lab’s own generated case identifier, year, state, county FIPS code, and all coded variables that appear in the published aggregates.

Access is granted on a case-by-case basis to researchers who need case-level data to replicate or extend published findings. To request access, contact the Lab through the corrections and feedback page with a brief description of your research purpose. Aggregate counts on this site remain the primary publication; the restricted deposit exists to support replication, not to substitute for it.

How small counts are protected

A count below five can point to an identifiable case. The suppression rules are therefore as follows.

First, counts of one to four are suppressed and shown as “suppressed” in the data table. A count of zero is displayed.

Second, suppressing one count is not sufficient on its own. Because the total for a chart is published, a single hidden count could be recovered by subtraction. The Lab therefore also conceals the next-smallest count, so that the hidden values are only recoverable as a combined total of at least five.

Third, the same rule applies within each row when a chart is split two ways — by year and by camera footage, for example — because yearly totals are published elsewhere on the site.

If one category in a chart held only three cases, that category would be suppressed together with the next-smallest category in the chart.

How the rules are enforced

The export program reads the research database in read-only mode and accesses only a short list of named columns; names, locations, narratives, and sources are not on that list. The program applies the suppression rules before anything is written. The website re-checks each snapshot when it loads and refuses to display a snapshot containing a forbidden field, an unsuppressed small count, or a count that could be recovered from published totals. Each snapshot records a fingerprint of the source data and of the published figures so that any figure can be traced to the exact data behind it.

Concerns and corrections

Anyone who believes a published figure is wrong, or who has a privacy concern, can use the process described on the corrections page.

How to cite

Edged-Weapon Response Lab. (2026). Suppression and privacy. In EWRL Methods Documentation (Version 1.0). https://droptheknife.org/methods/suppression-and-privacy