I've spent the past year building a database, and it's now public: Bound by Fear, an archive of abuse enabled by religious and spiritual authority, drawn entirely from court judgments and official findings.
It holds 1,283 records spanning 1819 to 2026 and roughly 60 countries. Of those, 1,028 meet a published inclusion test, and 95% of that set record criminal convictions. Every record cites its sources — 2,241 of them, each logged individually.
One rule, applied everywhere
A record qualifies when the victim was subject to the perpetrator's spiritual authority, and that authority is what made the harm possible or impossible to refuse.
That rule throws out more than it lets in. A devout parent who abuses their own child fails it, because the crime doesn't depend on the authority being spiritual. A cleric who commits ordinary fraud fails it. Religious persecution — where an accuser holds no authority over the victim — runs the relationship backwards, and is documented separately for exactly that reason. Convictions overturned on appeal are excluded by name, and I check appellate history on every record.
The point of a single rule is that it has to cut against cases I might want to keep. A criterion that bends for a famous name isn't a criterion, it's a preference.
What the data actually says
The breakdown surprises people who expect a project like this to be about fringe groups. Christian institutions account for 778 of the records. Islamic contexts account for 111, new religious movements 69, Buddhist 41, Jewish 34.
The argument the data supports isn't that any one tradition is worse. It's that the mechanism doesn't vary. Authority makes refusal impossible, and it does that in a Catholic parish, a madrasa, a Zen centre and an ashram in the same shape.
What it can't tell you
This is a convenience sample. It records where abuse was investigated, prosecuted, and reported in a form I could verify and cite. That is a much narrower thing than where abuse happens, and any reading of the country counts as a map of risk would be wrong. Places with functioning courts and a free press generate records; places without them don't.
The country figures are approximate for a second reason: location is a free-text field, not a normalized jurisdiction code. The coverage notes say all of this before they say anything else, which is deliberate. A dataset that won't tell you what it can't support isn't one you should trust.
Open data
Everything is downloadable as JSON and CSV, and there's a versioned read-only JSON API with an OpenAPI 3.1 specification and a published codebook. No account, no key, no rate limit.
Researchers and reporters are the audience I built the API for. If you want to check my work, the method is public, the corrections process is open, and classification changes are dated in a changelog — including a bias audit last week that moved several categories after I found the taxonomy was treating non-Western traditions differently from Christian ones.
If you find something wrong, tell me and I'll fix it in the open.
