Skip to data stories
Documentary record Bound by Fear

Data stories

Mechanisms, not traditions

Across 1,023 documented cases drawn from prosecutions, litigation and official investigations, the same handful of coercive mechanisms recur in belief systems that share no history, no theology and no geography. This page sets out what that corpus supports, what it cannot support, and where the evidence thins out.

Read this before the charts

This corpus is a convenience sample of 1,023 core records — documented, sourced instances of spiritual authority used as the mechanism of coercion. It is not a probability sample of abuse, of any tradition's practitioners, or of any population. It cannot support claims about how often this happens per capita, within any tradition, or relative to any tradition's size. What it can support is a comparison of documented cases against each other.

38% of core records fall outside Christian traditions. That is the neutral baseline used throughout: a mechanism spread evenly across the corpus would sit near 38%. Departures from it are what the charts make visible. Cells resting on fewer than 5 records are suppressed rather than printed as a small number.

One record is one prosecuted case, whether it harmed one person or thousands. The corpus carries no structured victim count, so every figure on this page counts cases, not victims. See “Every figure here counts cases, not people” below for what that means and what would fix it.

What this page cannot show

Every figure here counts cases, not people

Every number, bar, table cell and percentage on this page counts prosecuted cases. One record is one case, whether the underlying conduct harmed a single person or several hundred. A single-victim assault and a mass-casualty institutional-abuse case each contribute exactly one record to every chart above and below.

That has a direct and systematic consequence: mass-victim events are understated in every proportion this page shows. Institutional abuse, trafficking networks and other cases involving many victims under one prosecution count the same as a one-victim case in every rate, share and ranking above — the fatality ladder, the mechanism distribution, the harm-type breakdown, the lift matrix, all of it. None of those figures can be reweighted by victim count, because the underlying field does not exist.

This is not a gap this page can quietly work around. Only 22 of 1,023 core records (2%) state a victim count anywhere in their free text, and extraction cannot responsibly produce a victim-count figure at that coverage — publishing one would mean inventing a number for the other 98% of records. So none is shown: no "approximately N victims," no per-case average, no weighted total, anywhere on this page.

The fix is a structured countBand field (schema v2), populated from official sources at the point a record is added rather than mined after the fact from whatever free text happens to be on hand. The extraction pilot that scoped this measured a victim count stated in roughly 48% of official court and prosecutor documents — over twenty times the 2% this page can currently draw on from the existing corpus text. Populating countBand going forward, and back-filling it for existing sources where the underlying document supports it, is the single highest-value data addition on this project's roadmap: it is the only change that would let every chart on this page eventually show a victim-weighted view alongside the case-weighted one it shows now.

The lead finding

The mechanism, not the tradition, predicts whether a case ends in death

Some coercion mechanisms are far likelier than others to appear in a record that ended in death or ritual murder. Exorcism, ritual initiation and healing claims run an order of magnitude more lethal than pastoral authority or institutional obedience — a spread of roughly 35× from top to bottom. The obvious objection is that this is really a story about which traditions are violent. It is not: the same ordering holds within Christian records alone (the hollow markers), where a healing or exorcism claim is roughly ten times more likely to accompany a death than pastoral-clerical authority is.

Fatal-outcome rate by coercion mechanism, with the within-Christianity rate on the same rowA dot-and-bar ladder. For each mechanism, a solid bar and dot show the share of its records whose outcome was death or ritual murder, across all traditions; a hollow ring on the same row shows the same rate computed within Christian records only. The two orderings agree, so the mechanism rather than the tradition predicts lethality. Each row is labelled with the number of records. Full figures follow in the accompanying data table. 0% 15% 31% 46% 61% Violent occult ideology or sacrifice 61% n=18 Exorcism or deliverance 24% n=45 Prosperity, divination, or curse-removal claim 23% n=31 Prophecy or divine command 22% n=55 Ritual, oath, or initiation 19% n=100 Spiritual healing or treatment 19% n=134 Curse or witchcraft threat 12% n=41 Other spiritual authority or belief 12% n=59 Threatened spiritual consequence 12% n=85 Guru or spiritual-teacher authority 6% n=187 Sacred sex, purity, or sexual-energy claim 6% n=32 Religious marriage or family control 6% n=16 Institutional obedience or isolation 4% n=322 Pastoral or clerical authority 2% n=626
  • All traditions
  • Within Christian records only
Table: fatal-outcome rate by mechanism, overall and within Christianity
Share of each mechanism's records whose recorded outcome was death or ritual murder, across all traditions and within Christian records alone. Denominator for each rate is the record count in that column. Within-Christian rate is shown only where at least 5 Christian records carry the mechanism. Corpus-wide fatal rate is 7.3%; within Christianity it is 4.2%.
Mechanism Records (n) Fatal outcomes Fatal rate (all) Christian records (n) Fatal rate (Christian only)
Violent occult ideology or sacrifice 18 11 61.1% 1 < 5
Exorcism or deliverance 45 11 24.4% 24 25.0%
Prosperity, divination, or curse-removal claim 31 7 22.6% 4 < 5
Prophecy or divine command 55 12 21.8% 37 16.2%
Ritual, oath, or initiation 100 19 19.0% 27 3.7%
Spiritual healing or treatment 134 25 18.7% 41 31.7%
Curse or witchcraft threat 41 5 12.2% 11 27.3%
Other spiritual authority or belief 59 7 11.9% 12 8.3%
Threatened spiritual consequence 85 10 11.8% 47 17.0%
Guru or spiritual-teacher authority 187 12 6.4% 29 0.0%
Sacred sex, purity, or sexual-energy claim 32 2 6.3% 13 0.0%
Religious marriage or family control 16 1 6.3% 10 0.0%
Institutional obedience or isolation 322 13 4.0% 233 2.1%
Pastoral or clerical authority 626 11 1.8% 560 1.8%

Denominator: each mechanism's own record count (1,023 core records total). “Fatal” means the record's type is death or ritual murder. As everywhere on this page, these are prosecuted and documented cases only — the rate is the share of documented cases that were fatal, not a mortality risk for anyone subject to the mechanism. This is a CASE rate, not a victim rate: a 24% fatality rate among exorcism cases is a statement about cases, and one fatal case may involve one death or many. The within-Christian columns hold the tradition fixed to show the ordering is a property of the mechanism.

The structural finding

Most coercion mechanisms are structural, not tradition-bound

If these mechanisms were features of particular belief systems, each would cluster inside one tradition. Most do not. The dashed line marks the 38% corpus baseline; bars reaching well past it are mechanisms that recur across unrelated belief systems.

Share of each coercion mechanism occurring outside Christian traditionsA horizontal bar chart split into two bands. The first band holds well-evidenced mechanisms; the second holds four thinly evidenced mechanisms based on fewer than 33 records. Bar thickness increases with the number of records behind each estimate, and a horizontal whisker shows the 95 percent confidence interval. A dashed vertical line marks the 38 percent corpus baseline. Full figures follow in the accompanying data table. 38% BASELINEWell-evidenced mechanisms (n ≥ 33) Guru or spiritual-teacher authority 85% n=187 Other spiritual authority or belief 80% n=59 Curse or witchcraft threat 73% n=41 Ritual, oath, or initiation 73% n=100 Spiritual healing or treatment 69% n=134 Exorcism or deliverance 47% n=45 Threatened spiritual consequence 45% n=85 Prophecy or divine command 33% n=55 Institutional obedience or isolation 28% n=322 Pastoral or clerical authority 11% n=626 Thinly evidenced — read as indicative only (n < 33) Violent occult ideology or sacrifice 94% n=18 low n Prosperity, divination, or curse-removal claim 87% n=31 low n Sacred sex, purity, or sexual-energy claim 59% n=32 low n Religious marriage or family control 38% n=16 low n
Table: mechanism share outside Christian traditions
Share of each coercion mechanism's records occurring outside Christian traditions, with record count and 95% Wilson confidence interval. Denominator is each mechanism's own record count. Corpus baseline is 38%.
Mechanism Share outside Christian traditions 95% interval Records (n) Evidence strength
Violent occult ideology or sacrifice 94% 74%–99% 18 Low n — indicative only
Prosperity, divination, or curse-removal claim 87% 71%–95% 31 Low n — indicative only
Guru or spiritual-teacher authority 85% 79%–89% 187 Well evidenced
Other spiritual authority or belief 80% 68%–88% 59 Well evidenced
Curse or witchcraft threat 73% 58%–84% 41 Well evidenced
Ritual, oath, or initiation 73% 64%–81% 100 Well evidenced
Spiritual healing or treatment 69% 61%–77% 134 Well evidenced
Sacred sex, purity, or sexual-energy claim 59% 42%–75% 32 Low n — indicative only
Exorcism or deliverance 47% 33%–61% 45 Well evidenced
Threatened spiritual consequence 45% 35%–55% 85 Well evidenced
Religious marriage or family control 38% 18%–61% 16 Low n — indicative only
Prophecy or divine command 33% 22%–46% 55 Well evidenced
Institutional obedience or isolation 28% 23%–33% 322 Well evidenced
Pastoral or clerical authority 11% 8%–13% 626 Well evidenced

Denominator: each mechanism's own record count, within 1,023 core records. Bar thickness encodes n; the whisker is a 95% Wilson interval. The four mechanisms in the lower band rest on 16–32 records — their position is real but unstable, and a handful of new records could move them substantially. They are banded separately so they are never ranked against estimates built on hundreds of records.

Breadth

The broadest mechanisms span more than a dozen unrelated traditions

A mechanism appearing in one or two traditions may be a local practice. One appearing across a dozen unrelated traditions is a property of spiritual authority as a structure. Distinct-tradition count is a lower bound, and it is sensitive to how finely this taxonomy splits traditions.

Number of distinct traditions each coercion mechanism appears in, with the mechanism's record count. Denominator is 1,023 core records.
Mechanism Distinct traditions Records (n)
Ritual, oath, or initiation 17 100
Other spiritual authority or belief 14 59
Spiritual healing or treatment 14 134
Guru or spiritual-teacher authority 13 187
Prosperity, divination, or curse-removal claimlow n 12 31
Threatened spiritual consequence 12 85
Institutional obedience or isolation 10 322
Exorcism or deliverance 9 45
Violent occult ideology or sacrificelow n 8 18
Curse or witchcraft threat 8 41
Sacred sex, purity, or sexual-energy claimlow n 8 32
Pastoral or clerical authority 8 626
Religious marriage or family controllow n 5 16
Prophecy or divine command 5 55

The complication

Compound coercion is the majority case, not the exception

603 of 1,023 core records — 59% — name two or more mechanisms operating at once, up to 6 on a single record. Treating these mechanisms as alternatives misreads the corpus: in most documented cases they are layered, an authority claim reinforced by a threatened spiritual consequence, or a healing claim wrapped inside institutional obedience.

Number of coercion mechanisms named per recordA bar chart showing how many core records name one, two, three or more distinct coercion mechanisms. Figures follow in the accompanying table. 1 mechanism 420 41% 2 mechanisms 498 49% 3 mechanisms 91 9% 4 mechanisms 9 1% 5 mechanisms 4 0% 6 mechanisms 1 0%
Table: mechanisms named per record, and whether stacking predicts severity
Distribution of the number of coercion mechanisms named per record, with conviction rate and median parsed custodial sentence for each band. Denominator for share is 1,023 core records; sentence figures use only the subsample whose outcome text names an explicit term.
Mechanisms named Records Share of corpus
1 420 41.1%
2 498 48.7%
3 91 8.9%
4 9 0.9%
5 4 0.4%
6 1 0.1%
Fatal-outcome rate by number of mechanisms named. Fatal means record type death or ritual murder; denominator is the band's record count.
Band Records Fatal outcomes Fatal rate
One mechanism 420 22 5.2%
Two mechanisms 498 39 7.8%
Three or more 105 14 13.3%

Stacked coercion is more often fatal — but the gap attenuates with documentation

Records naming more mechanisms are more often fatal: one mechanism 5.2%, two mechanisms 7.8%, three or more 13.3%. The gradient is real and survives adjustment for how heavily each record is documented (source-adjusted odds ratio for two-or-more versus one mechanism ≈ 1.75). But it attenuates: among the most heavily documented records the gap nearly closes, so this is a gradient that survives adjustment with attenuation, not a clean law. We do not chart conviction rate (circular — inclusion is conditioned on prosecution) or sentence length (free text mixes concurrent, cumulative and life terms and is not comparable across jurisdictions).

Denominator: 1,023 core records. Fatal-outcome rate is the share of records whose type is death or ritual murder — a documented-outcome measure, not a risk. Stacking is associational: heavily coercive situations may simply generate more of everything, including documentation.

Which mechanisms stack

Authority claims are what the other mechanisms attach to

Each cell counts the records on which two mechanisms appear together. Darker and larger numbers mean heavier co-occurrence; every cell also prints its count, so nothing here depends on reading a shade. Cells below 5 records are suppressed and left blank.

This chart is wider than the screen. Scroll horizontally within it to see all mechanisms, or use the data table below instead.
Co-occurrence matrix of coercion mechanismsA symmetric grid in which each cell shows how many core records name both the row mechanism and the column mechanism. Counts are printed inside every unsuppressed cell. Full figures follow in the accompanying data table.Violent occult ideology 11 Prosperity, divination, 6 7 7 Guru or spiritual-teache 21 17 13 18 7 63 37Other spiritual authorit 12 7 5Curse or witchcraft thre 6 5 23 12 8Ritual, oath, or initiat 11 7 21 5 15 7 5 6 10 25Spiritual healing or tre 7 17 12 23 15 17 19 9 9 32Sacred sex, purity, or s 13 7 5 7 9 10Exorcism or deliverance 12 17 7 16Threatened spiritual con 18 5 19 5 7 15 16 31Religious marriage or fa 6 5Prophecy or divine comma 7 6 9 7 15 6 14 13Institutional obedience 63 7 10 9 9 16 14 214Pastoral or clerical aut 37 5 8 25 32 10 16 31 5 13 214 Violent occult ideology Prosperity, divination, Guru or spiritual-teacheOther spiritual authoritCurse or witchcraft threRitual, oath, or initiatSpiritual healing or treSacred sex, purity, or sExorcism or deliveranceThreatened spiritual conReligious marriage or faProphecy or divine commaInstitutional obedience Pastoral or clerical aut
Table: mechanism co-occurrence counts
Number of core records on which each pair of coercion mechanisms appears together. Blank cells are suppressed because fewer than 5 records support them. Denominator is 1,023 core records.
Mechanism Violent occult ideology or sacrificeProsperity, divination, or curse-removal claimGuru or spiritual-teacher authorityOther spiritual authority or beliefCurse or witchcraft threatRitual, oath, or initiationSpiritual healing or treatmentSacred sex, purity, or sexual-energy claimExorcism or deliveranceThreatened spiritual consequenceReligious marriage or family controlProphecy or divine commandInstitutional obedience or isolationPastoral or clerical authority
Violent occult ideology or sacrifice — <5 <5 <5 <5 11 <5 <5 <5 <5 <5 <5 <5 <5
Prosperity, divination, or curse-removal claim <5 — <5 <5 6 7 7 <5 <5 <5 <5 <5 <5 <5
Guru or spiritual-teacher authority <5 <5 — <5 <5 21 17 13 <5 18 <5 7 63 37
Other spiritual authority or belief <5 <5 <5 — <5 <5 12 <5 <5 <5 <5 <5 7 5
Curse or witchcraft threat <5 6 <5 <5 — 5 23 <5 12 <5 <5 <5 <5 8
Ritual, oath, or initiation 11 7 21 <5 5 — 15 7 <5 5 <5 6 10 25
Spiritual healing or treatment <5 7 17 12 23 15 — <5 17 19 <5 9 9 32
Sacred sex, purity, or sexual-energy claim <5 <5 13 <5 <5 7 <5 — <5 5 <5 7 9 10
Exorcism or deliverance <5 <5 <5 <5 12 <5 17 <5 — 7 <5 <5 <5 16
Threatened spiritual consequence <5 <5 18 <5 <5 5 19 5 7 — <5 15 16 31
Religious marriage or family control <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 — 6 <5 5
Prophecy or divine command <5 <5 7 <5 <5 6 9 7 <5 15 6 — 14 13
Institutional obedience or isolation <5 <5 63 7 <5 10 9 9 <5 16 <5 14 — 214
Pastoral or clerical authority <5 <5 37 5 8 25 32 10 16 31 5 13 214 —

Denominator: 1,023 core records. Cells are counts of records, not of victims or defendants. The matrix is symmetric; the diagonal is undefined and shown blank.

The deeper cut

When an institution is named, the coercion changes shape

645 of 1,023 core records name at least one organization; 378 name none. Those two populations do not look alike. Institutional obedience appears in 43% of records naming an organization against 12% of those that do not — the single largest divergence in the corpus. Where no institution is named, healing and curse claims rise instead: coercion that has to be manufactured case by case rather than inherited from a standing hierarchy.

Mechanism prevalence in records naming an institution versus records naming noneA paired bar chart. For each mechanism, the upper solid bar is its prevalence among records that name an organization and the lower outlined bar is its prevalence among records that do not. Each bar is labelled with its percentage. Full figures follow in the accompanying data table. Institutional obedience or isolation 43% inst. 12% none Pastoral or clerical authority 70% inst. 46% none Spiritual healing or treatment 8% inst. 22% none Ritual, oath, or initiation 7% inst. 14% none Curse or witchcraft threat 1% inst. 8% none Exorcism or deliverance 2% inst. 8% none Other spiritual authority or belief 4% inst. 9% none Guru or spiritual-teacher authority 17% inst. 20% noneInstitution namedNo institution named
Table: mechanism prevalence by whether an institution is named
Prevalence of each coercion mechanism among the 645 core records naming an organization, against the 378 that name none. Percentages use each population as its own denominator, so they do not sum to 100.
Mechanism Institution named (n) Institution named (%) No institution (n) No institution (%) Difference
Institutional obedience or isolation 276 42.8% 46 12.2% +30.6%
Pastoral or clerical authority 451 69.9% 175 46.3% +23.6%
Spiritual healing or treatment 52 8.1% 82 21.7% −13.6%
Ritual, oath, or initiation 46 7.1% 54 14.3% −7.2%
Curse or witchcraft threat 9 1.4% 32 8.5% −7.1%
Exorcism or deliverance 16 2.5% 29 7.7% −5.2%
Other spiritual authority or belief 26 4.0% 33 8.7% −4.7%
Guru or spiritual-teacher authority 110 17.1% 77 20.4% −3.3%
Prosperity, divination, or curse-removal claim 12 1.9% 19 5.0% −3.2%
Prophecy or divine command 41 6.4% 14 3.7% +2.7%
Sacred sex, purity, or sexual-energy claim 26 4.0% 6 1.6% +2.4%
Threatened spiritual consequence 48 7.4% 37 9.8% −2.3%
Religious marriage or family control 14 2.2% 2 0.5% +1.6%
Violent occult ideology or sacrifice 12 1.9% 6 1.6% +0.3%

Concentration among named organizations is weak. 709 distinct organizations appear across 645 records, and only 11 are named on 3 or more. This corpus does not show a small number of institutions accounting for the bulk of documented cases; it shows a long tail. That is at least partly a documentation artifact — large denominations are indexed, reported and litigated under a consistent name, while a standalone congregation may appear once and never again.

Organizations named on 3 or more core records, with record counts. Denominator is the 645 records naming an organization.
Organization named on 3 or more records Records
Church of Jesus Christ of Latter-day Saints 7
Diocese of Lansing 5
Southern Baptist Convention 5
764 4
Missionary Oblates of Mary Immaculate 4
'Two by Twos' / 'The Truth' (nameless sect) 3
Archdiocese of Detroit 3
Church of England 3
Congregation of Christian Brothers 3
Jesus Army / Jesus Fellowship Church 3
Marist Brothers 3

Denominator: 645 records naming an organization and 378 naming none, of 1,023 core records. Organization names are recorded as they appear in sources and are not deduplicated across dioceses, provinces or renamings, so counts here are a floor.

What we did not chart

The paper trail runs deep, but three obvious charts would have lied

The court record behind these cases is substantial: 277 records carry appellate history, and 415 carry proceeding text naming an explicit custodial term. That depth invites three charts we deliberately did not build, because each would have manufactured a finding the data cannot support.

  • Conviction rate by mechanism. Inclusion in this corpus is conditioned on a case having been prosecuted or officially found — so nearly every record is a conviction by construction. There is no acquittal denominator, which makes any conviction-rate comparison circular.
  • Sentence severity by mechanism. The 415 parsed terms mix concurrent, cumulative and life sentences in free text across jurisdictions with fundamentally different sentencing regimes. They are not on a common scale and cannot be compared.
  • Time from offence to prosecution. The dating field records the year of prosecution or official finding, not the year of the underlying conduct, so any computed “lag” collapses to zero as an artifact rather than a measurement.

Every comparison on this page is also internal: mechanism against mechanism, inside one corpus, under one inclusion criterion. There is no external control — no matched sample of abuse prosecutions without a spiritual-authority element (teachers, coaches, doctors, youth leaders) to benchmark against. So the page can say exorcism cases are roughly 35 times more often fatal than pastoral cases, but not whether either is high or low against secular authority abuse. The planned control is a matched-methodology secular comparison corpus — same conviction-must-stand criterion, same coding, minus the spiritual element. No external statistic (NCANDS, BJS, ONS) is a valid substitute: their selection differs at every stage — victims rather than cases, substantiated rather than prosecuted, incidence rather than a curated archive.

Mechanism and harm

Different mechanisms attach to different harms

Each row shows how a mechanism's own records distribute across recorded harm types. Percentages are within each mechanism, so a mechanism resting on 626 records and one resting on a few dozen can be compared down a column. Counts below 5 are suppressed.

Distribution of each coercion mechanism's records across recorded harm types, shown as counts. Cells below 5 records are suppressed. Denominator for each row is that mechanism's own record count.
Mechanism n Child abuseSexual abuseDeathTraffickingRitual murderFraudPhysical abuseInstitutional abuseRitual crimeInterventionHonor killingForced marriageRetaliation
Pastoral or clerical authority 626 367 (59%) 223 (36%) 11 (2%) <5 <5 <5 5 (1%) 12 (2%) <5 <5 <5 <5 <5
Institutional obedience or isolation 322 163 (51%) 108 (34%) 13 (4%) 8 (2%) <5 <5 11 (3%) 15 (5%) <5 <5 <5 <5 <5
Guru or spiritual-teacher authority 187 100 (53%) 64 (34%) 11 (6%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5
Spiritual healing or treatment 134 18 (13%) 85 (63%) 20 (15%) <5 5 (4%) <5 <5 <5 <5 <5 <5 <5 <5
Ritual, oath, or initiation 100 38 (38%) 30 (30%) 7 (7%) 8 (8%) 12 (12%) <5 <5 <5 <5 <5 <5 <5 <5
Threatened spiritual consequence 85 24 (28%) 43 (51%) 10 (12%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5
Other spiritual authority or belief 59 18 (31%) 21 (36%) 5 (8%) 5 (8%) <5 5 (8%) <5 <5 <5 <5 <5 <5 <5
Prophecy or divine command 55 18 (33%) 20 (36%) 9 (16%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5
Exorcism or deliverance 45 8 (18%) 23 (51%) 7 (16%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5
Curse or witchcraft threat 41 7 (17%) 22 (54%) <5 <5 <5 5 (12%) <5 <5 <5 <5 <5 <5 <5
Sacred sex, purity, or sexual-energy claim 32 6 (19%) 24 (75%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5
Prosperity, divination, or curse-removal claim 31 <5 8 (26%) 6 (19%) <5 <5 12 (39%) <5 <5 <5 <5 <5 <5 <5
Violent occult ideology or sacrifice 18 7 (39%) <5 5 (28%) <5 6 (33%) <5 <5 <5 <5 <5 <5 <5 <5
Religious marriage or family control 16 8 (50%) <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5

Denominator: each mechanism's own record count. Harm type is a single primary classification per record, so a case involving several harms is counted once under its recorded type — these rows understate co-occurring harm. The taxonomy's tail is thin: 5 of the 13 recorded types rest on fewer than 5 records corpus-wide, so most columns in this table are suppressed for most mechanisms. The type distribution and the mechanism-type lift below handle that thinness directly.

Type of crime

Child abuse and sexual abuse dominate the recorded harm types; most of the rest is a thin tail

Across 1,023 core records, child abuse (49.1%) and sexual abuse (35.3%) together account for the large majority of recorded harm types. Nine of the thirteen recorded types rest on fewer than 25 records, and 5 of them fall below this page's 5-record suppression floor entirely (8 records combined). Those are not drawn as individual bars — a chart that puts a single-record type on the same axis as a 499-record type invites readers to compare them as equally supported estimates, which they are not. They are pooled into one labelled band instead.

Distribution of core records by recorded harm typeA horizontal bar chart of 9 rows, one per recorded harm type plus a pooled "other" band for types below the suppression floor, showing each type's share of 1,023 core records. Full figures follow in the accompanying data table. Child abuse 502 (49.1%) Sexual abuse 361 (35.3%) Death 52 (5.1%) Trafficking 25 (2.4%) Ritual murder 23 (2.2%) Fraud 18 (1.8%) Physical abuse 18 (1.8%) Institutional abuse 16 (1.6%) Other harm types 8 (0.8%)
Table: recorded harm type distribution
Distribution of core records by recorded primary harm type. Types below 5 records corpus-wide are pooled into a single "other" row rather than shown individually. Denominator is 1,023 core records.
Harm type Records (n) Share of corpus
Child abuse 502 49.1%
Sexual abuse 361 35.3%
Death 52 5.1%
Trafficking 25 2.4%
Ritual murder 23 2.2%
Fraud 18 1.8%
Physical abuse 18 1.8%
Institutional abuse 16 1.6%
Other harm types 8 0.8%

Pooled “other” band: Ritual crime (4); Intervention (1); Honor killing (1); Forced marriage (1); Retaliation (1).

Denominator: 1,023 core records. Type is a single primary classification per record. This corpus's taxonomy tail is thin — several categories (ritual crime, intervention, honor killing, forced marriage, retaliation) exist as recognised classifications but are each documented in this archive on a handful of cases or fewer, which is a property of what this convenience sample was able to source, not a claim about how rare those harms are.

What each mechanism is for

Mechanisms are not interchangeable routes to a generic harm — each is fused to a characteristic crime

A raw mechanism-by-type crosstab mostly re-prints the size distribution above: child abuse is 49% of the corpus, so it dominates every row regardless of mechanism. Lift corrects for that: lift = observed ÷ expected, where expected assumes a mechanism's records split across harm types exactly as the corpus overall does. Lift near 1.0 means a pairing is unremarkable; departures show what a mechanism is actually for. Prosperity-divination is a fraud instrument (22.0× expected). Violent occult ideology is a killing instrument (14.8× expected on ritual murder). Pastoral-clerical authority runs the other way — heavily over-represented in child abuse elsewhere on this page, but under-represented here for death: it is the mechanism most likely to leave a victim alive to testify, which is part of why it dominates this corpus's headline count at all.

Mechanism-type pairings that depart most from their expected countA diverging bar chart on a log scale centred at lift = 1 (no difference from the corpus baseline). Bars extending right show a mechanism-type pairing occurring more often than expected; bars extending left show one occurring less often than expected. Every bar is labelled with its lift multiple and its record count n. Only pairings resting on at least 5 records are shown. Full figures follow in the accompanying data table.1× (expected) Prosperity, divination, or curse-removal claim × Fraud 22.0× n=12 Violent occult ideology or sacrifice × Ritual murder 14.8× n=6 Curse or witchcraft threat × Fraud 6.9× n=5 Violent occult ideology or sacrifice × Death 5.5× n=5 Ritual, oath, or initiation × Ritual murder 5.3× n=12 Other spiritual authority or belief × Fraud 4.8× n=5 Prosperity, divination, or curse-removal claim × Death 3.8× n=6 Spiritual healing or treatment × Child abuse 0.3× n=18 Other spiritual authority or belief × Trafficking 3.5× n=5 Ritual, oath, or initiation × Trafficking 3.3× n=8 Prophecy or divine command × Death 3.2× n=9 Exorcism or deliverance × Death 3.1× n=7 Institutional obedience or isolation × Institutional abuse 3.0× n=15 Spiritual healing or treatment × Death 2.9× n=20 Pastoral or clerical authority × Death 0.3× n=11 Curse or witchcraft threat × Child abuse 0.3× n=7 Exorcism or deliverance × Child abuse 0.4× n=8 Sacred sex, purity, or sexual-energy claim × Child abuse 0.4× n=6 Threatened spiritual consequence × Death 2.3× n=10 Pastoral or clerical authority × Physical abuse 0.5× n=5 Sacred sex, purity, or sexual-energy claim × Sexual abuse 2.1× n=24 Institutional obedience or isolation × Physical abuse 1.9× n=11 Spiritual healing or treatment × Sexual abuse 1.8× n=85 Threatened spiritual consequence × Child abuse 0.6× n=24 Other spiritual authority or belief × Death 1.7× n=5 Spiritual healing or treatment × Ritual murder 1.7× n=5 Other spiritual authority or belief × Child abuse 0.6× n=18 Curse or witchcraft threat × Sexual abuse 1.5× n=22
  • Over-represented (lift > 1)
  • Under-represented (lift < 1)
Table: mechanism-type lift, all pairings at or above 5 records
Observed versus expected record count for each mechanism-type pairing with n ≥ 5. Expected assumes the mechanism's records distribute across types exactly as the corpus overall does. Lift = observed ÷ expected. Denominator for each mechanism's expected count is that mechanism's own record count; type shares are corpus-wide.
Mechanism Type n Expected Lift
Prosperity, divination, or curse-removal claim Fraud 12 0.5 22.00×
Violent occult ideology or sacrifice Ritual murder 6 0.4 14.83×
Curse or witchcraft threat Fraud 5 0.7 6.93×
Violent occult ideology or sacrifice Death 5 0.9 5.46×
Ritual, oath, or initiation Ritual murder 12 2.2 5.34×
Other spiritual authority or belief Fraud 5 1.0 4.82×
Prosperity, divination, or curse-removal claim Death 6 1.6 3.81×
Spiritual healing or treatment Child abuse 18 65.8 0.27×
Other spiritual authority or belief Trafficking 5 1.4 3.47×
Ritual, oath, or initiation Trafficking 8 2.4 3.27×
Prophecy or divine command Death 9 2.8 3.22×
Exorcism or deliverance Death 7 2.3 3.06×
Institutional obedience or isolation Institutional abuse 15 5.0 2.98×
Spiritual healing or treatment Death 20 6.8 2.94×
Pastoral or clerical authority Death 11 31.8 0.35×
Curse or witchcraft threat Child abuse 7 20.1 0.35×
Exorcism or deliverance Child abuse 8 22.1 0.36×
Sacred sex, purity, or sexual-energy claim Child abuse 6 15.7 0.38×
Threatened spiritual consequence Death 10 4.3 2.31×
Pastoral or clerical authority Physical abuse 5 11.0 0.45×
Sacred sex, purity, or sexual-energy claim Sexual abuse 24 11.3 2.13×
Institutional obedience or isolation Physical abuse 11 5.7 1.94×
Spiritual healing or treatment Sexual abuse 85 47.3 1.80×
Threatened spiritual consequence Child abuse 24 41.7 0.58×
Other spiritual authority or belief Death 5 3.0 1.67×
Spiritual healing or treatment Ritual murder 5 3.0 1.66×
Other spiritual authority or belief Child abuse 18 29.0 0.62×
Curse or witchcraft threat Sexual abuse 22 14.5 1.52×

Formula: lift = observed ÷ expected, where expected = (mechanism's record count) × (that type's share of all 1,023 core records), i.e. the count predicted if the mechanism's harm types were distributed exactly like the corpus overall. Every cell shown rests on at least 5 observed records; n is printed alongside every lift value so a 22× lift on n=12 reads as more supported than a 3.5× lift on n=5. This is co-occurrence within prosecuted cases, not a causal or prevalence claim, and every count is a case count, not a victim count.

Sentencing

A sentence is stated in the record for 41% of core cases — and most of the time it is not

Of 1,023 core records, 63 state an explicit life sentence, 352 state a numeric custodial term, and 608 state no sentence at all in the text this archive could source. That leaves 415 of 1,023 records (41%) with any sentencing information whatsoever. The 608 "not stated" is itself a finding, not a gap to be filled: it describes what court and press reporting actually captures for a case, not what any court necessarily imposed.

This page does not chart a median or mean sentence length. A prior analysis pass confirmed the 352 numeric terms are not normalizable into a single comparable figure: the free text mixes concurrent terms, cumulative terms stacked across separate counts, and open-ended "X years to life" phrasing, drawn from jurisdictions with incompatible sentencing regimes — one parse of the underlying text reached 658 years on a single record, which is a real quotation, not a usable statistic. Inventing an average from numbers that are not on a common scale would manufacture a precision the source text does not have, so none is shown. What survives as a defensible chart is the three-way split above, and a binary — whether a life sentence was named at all — below.

Sentencing information stated in the record, by categoryA bar chart with three rows: records stating a life sentence, records stating a numeric custodial term, and records stating no sentence at all. Denominator is 1,023 core records. Full figures follow in the accompanying data table. Life sentence stated 63 (6%) Numeric term stated 352 (34%) No sentence stated 608 (59%)
Table: sentencing information stated, and life sentences by mechanism and type
Three-way split of sentencing information stated in the record. Denominator is 1,023 core records.
Category Records (n) Share of core records
Life sentence stated 63 6.2%
Numeric term stated 352 34.4%
No sentence stated 608 59.4%
Share of each mechanism's own records stating an explicit life sentence. Denominator for each row is that mechanism's own record count (not the 415-record stated-sentence subsample). Rows below 5 life-sentence records are omitted.
Mechanism Records (n) Life sentences Life-sentence rate
Violent occult ideology or sacrifice 18 5 27.8%
Prophecy or divine command 55 11 20.0%
Exorcism or deliverance 45 5 11.1%
Other spiritual authority or belief 59 6 10.2%
Guru or spiritual-teacher authority 187 16 8.6%
Threatened spiritual consequence 85 7 8.2%
Spiritual healing or treatment 134 10 7.5%
Ritual, oath, or initiation 100 6 6.0%
Pastoral or clerical authority 626 32 5.1%
Institutional obedience or isolation 322 9 2.8%
Share of each recorded harm type's own records stating an explicit life sentence. Denominator for each row is that type's own record count. Rows below 5 life-sentence records are omitted.
Harm type Records (n) Life sentences Life-sentence rate
Death 52 14 26.9%
Sexual abuse 361 21 5.8%
Child abuse 502 24 4.8%

The three-way split's denominator is 1,023 core records. Every other sentencing figure on this page — the 415-record stated-sentence subsample and the 41% figure above — is of the 415 records where a sentence is stated, never of the full 1,023. Life-sentence rate by mechanism and by type is a binary (was a life sentence named, yes or no) rather than a parsed year count, because a binary phrase match is far more robust at n=63 than any arithmetic on incomparable free text. Cells resting on fewer than 5 life-sentence records are omitted from both tables, consistent with the suppression rule used throughout this page.

The limits

This corpus cannot tell you whether abuse is rising

84% of all core records fall in the 2010s and 2020s. Plotted on a time axis that looks like a steep and alarming increase. It is not a finding. It is the shape of what a research project working today can locate, verify and source: court records digitise, press archives survive unevenly, limitation periods expire, and the further back you look the less there is to find. The chart below is therefore labelled as what it actually measures — this archive's research coverage — and no mechanism is drawn as a count over time anywhere on this page.

This chart is wider than the screen and starts scrolled to the decades holding the most records. Scroll horizontally to see earlier decades, or use the data table below instead.
Records held by decade of prosecution, on a year-true axis, with markers for events that expanded documentation — a measure of research coverage, not of incidenceA bar chart of how many core records this archive holds for each decade, positioned on a continuous year axis so gaps between decades are preserved, rising steeply toward the present. Dashed vertical lines mark events that made reporting, prosecution or documentation newly possible; each is labelled with its year and a short description. This reflects the availability of documentation, not the frequency of abuse. Full figures and the event list follow in the accompanying data table. 0 140 280 420 560 1810 1920 1940 1950 1960 1970 1980 1990 2000 2010 2020 1974 US CAPTA reporting law 1985 Gauthe clergy prosecution 2002 Boston Globe Spotlight 2009 Ireland Ryan Report 2013 Australian Royal Commissi… 2017 #MeToo 2018 PA grand jury; Oba revoke… 2022 Abe killing → Unification…
Table: records held by decade, and events that expanded documentation
Number of core records this archive holds for each decade, by decade of prosecution or official finding. This measures documentation availability and research coverage, not the incidence of abuse. Denominator is 1,023 core records.
Decade Records held Share of corpus
1810s 1 0.1%
1920s 1 0.1%
1940s 5 0.5%
1950s 8 0.8%
1960s 6 0.6%
1970s 4 0.4%
1980s 12 1.2%
1990s 51 5.0%
2000s 79 7.7%
2010s 296 28.9%
2020s 560 54.7%
Events marked on the chart. Each expanded the reporting, prosecution or documentation of this abuse; none is a cause of abuse. They are the documentation-opening events this corpus's records most visibly respond to.
Year Event affecting documentation
1974 US CAPTA reporting law
1985 Gauthe clergy prosecution
2002 Boston Globe Spotlight
2009 Ireland Ryan Report
2013 Australian Royal Commission
2017 #MeToo
2018 PA grand jury; Oba revokes juju oaths
2022 Abe killing → Unification Church reckoning

Denominator: 1,023 core records, dated by prosecution or official finding rather than by when the underlying conduct occurred. Markers show events that expanded reporting, prosecution or documentation — not causes of abuse. The rise in records tracks society's ability to see, and our research coverage, more than underlying incidence. Documentation-opening events occur in every region; these are the ones this corpus's records most visibly respond to, and the list is inevitably Western- and Anglophone-heavy. Read this chart as a description of the archive, not of the world.

The corpus also drifted as it was assembled

A second sampling artifact sits inside the first. This corpus was built in phases, and as the research strategy shifted the records changed character. Sorted by the order in which they were added, the earliest quartile is 16.1% fatal and 40% pastoral-clerical, while the most recent quartile is 0.8% fatal and 81% pastoral-clerical. The corpus blends two sampling regimes, and any pooled statistic on this page inherits that blend. We disclose it rather than hide it.

Fatal-outcome rate and pastoral-clerical share by collection-order quartile, from earliest-added records to most recent. Denominator for each rate is the quartile's record count.
Collection-order quartile Records Fatal rate Pastoral-clerical share
Q1 (earliest added) 255 16.1% 40.0%
Q2 255 4.3% 64.3%
Q3 255 8.2% 58.8%
Q4 (most recent) 258 0.8% 81.4%

Denominator: 1,023 core records split into four equal groups by addition order. This drift survives controlling for the decade of prosecution, so it is a property of how the archive was built, not merely of which decades it covers. The single highest-value future addition is a court-established victim-count band per record, which would let this page show victim-weighted views alongside the case-weighted ones it shows now; it is a provenance-gated task, not estimated here. Separately, a grooming or “position-of-trust” modus operandi appears in roughly 46% of records and cuts across all fourteen pretext categories — an orthogonal axis this taxonomy does not yet capture.

What can be compared across decades is the mix, not the volume

Composition is robust to the coverage problem in a way that counts are not. Normalizing each decade to 100% asks a different and answerable question: when this archive does hold a case from a given decade, what kind of coercion does it involve? Decades holding fewer than 5 records are omitted entirely rather than drawn as noise.

This chart is wider than the screen. Scroll horizontally within it to see every decade, or use the data table below instead.
Mechanism mix within each decade, normalized to 100 percentA one hundred percent stacked bar chart. Each decade's column is divided into bands showing what share of that decade's mechanism mentions each mechanism accounts for. Because every column sums to the same height, this shows composition and deliberately conveys nothing about volume. Full figures follow in the accompanying data table. 100% 75% 50% 25% 0% 100 1940s n=5 50 50 1950s n=8 56 44 1980s n=12 36 28 13 16 1990s n=51 40 24 17 2000s n=79 34 20 21 2010s n=296 37 16 12 22 2020s n=560
  • Pastoral or clerical authority
  • Institutional obedience or isolation
  • Guru or spiritual-teacher authority
  • Spiritual healing or treatment
  • Ritual, oath, or initiation
  • All other mechanisms
Table: mechanism mix by decade
Share of each decade's mechanism mentions accounted for by each mechanism, normalized to 100% within the decade. The denominator is mechanism mentions, not records, because a record may name several mechanisms. Decades holding fewer than 5 records are excluded.
Decade Records Mentions Pastoral or clerical authorityInstitutional obedience or isolationGuru or spiritual-teacher authoritySpiritual healing or treatmentRitual, oath, or initiationAll other mechanisms
1940s 5 5 100.0% (5)0.0% (0)0.0% (0)0.0% (0)0.0% (0)0.0% (0)
1950s 8 10 50.0% (5)50.0% (5)0.0% (0)0.0% (0)0.0% (0)0.0% (0)
1980s 12 16 56.3% (9)43.8% (7)0.0% (0)0.0% (0)0.0% (0)0.0% (0)
1990s 51 69 36.2% (25)27.5% (19)13.0% (9)7.2% (5)0.0% (0)15.9% (11)
2000s 79 127 40.2% (51)23.6% (30)5.5% (7)7.9% (10)5.5% (7)17.3% (22)
2010s 296 527 34.3% (181)19.9% (105)10.6% (56)6.5% (34)7.6% (40)21.1% (111)
2020s 560 935 37.0% (346)15.6% (146)11.6% (108)8.9% (83)5.1% (48)21.8% (204)

Denominator: mechanism mentions within each decade, normalized to 100%. Column heights are deliberately equal — this chart carries no information about volume, and none should be inferred from it. Decade totals are printed beneath each column so the thinness of the early decades stays visible.

Check it yourself

Dimensions and metrics explorer

Pick a dimension and a metric. Cells resting on fewer than 5 records are suppressed, never shown as a small number. This table renders fully without JavaScript; the controls change which slice it shows.

Showing Count by Tradition: 22 rows, of which 7 are suppressed for resting on fewer than 5 records.

Count by Tradition, across 1,023 core records. Rows resting on fewer than 5 records are marked suppressed.
Tradition Count
christian 637
islam 98
nrm 58
spiritual-healing 54
buddhism 33
judaism 27
hinduism 23
juju 17
new-age 17
indigenous-spiritual 11
afro-brazilian 11
muti 6
santeria 6
sikhism 5
western-occult 5
palo < 5 (suppressed)
haitian-vodou < 5 (suppressed)
fraud-us < 5 (suppressed)
left-hand-path < 5 (suppressed)
jainism < 5 (suppressed)
fraternal < 5 (suppressed)
northern-traditional < 5 (suppressed)

Findings

What the corpus supports

Most mechanisms are structural, not tradition-bound

Of 14 pretext categories, 11 classify as structural (spread evenly across many traditions) and 3 as tradition-bound, using the crossTraditionIndex >= 0.6 and distinctTraditions >= 5 rule.

n = 1,751 of 14 pretext categories

Caveat: This corpus is a convenience sample of successful, publicly documented prosecutions and official findings — not a random or representative sample of religious-authority abuse. It cannot support prevalence, incidence, or per-capita claims about any tradition, country, or mechanism. Absence of a record is not evidence of absence of abuse; see research/ANALYTICS.md and static/data/COVERAGE.md.

Conviction rates hold up across every mechanism

Conviction rate among core records varies by pretext mechanism from 84.8% (other-spiritual-authority) to 100.0% (curse-witchcraft), among mechanisms with n >= 5.

n = 1,751 of per-mechanism n, see supportingNumbers

Caveat: This corpus is a convenience sample of successful, publicly documented prosecutions and official findings — not a random or representative sample of religious-authority abuse. It cannot support prevalence, incidence, or per-capita claims about any tradition, country, or mechanism. Absence of a record is not evidence of absence of abuse; see research/ANALYTICS.md and static/data/COVERAGE.md. Conviction rate here reflects both case strength AND which mechanisms happen to be pursued to verified conviction in this dataset's source jurisdictions — it is not a measure of how convincing an accusation of that mechanism is in general.

The mix of mechanisms has shifted over time

Comparing records from decades 1810s-1970s (n=25) against 1980s-2020s (n=998), the pretext category with the largest share shift is institutional-obedience (60% -> 30.8% of records in each period).

25 early-period core records vs 998 late-period core records

Caveat: This corpus is a convenience sample of successful, publicly documented prosecutions and official findings — not a random or representative sample of religious-authority abuse. It cannot support prevalence, incidence, or per-capita claims about any tradition, country, or mechanism. Absence of a record is not evidence of absence of abuse; see research/ANALYTICS.md and static/data/COVERAGE.md. A temporal shift in documented-record share can reflect a change in what researchers were able to find and verify (search coverage improves over time, digitization of press archives is uneven by decade) as much as a change in underlying behavior.

Case documentation concentrates in a handful of countries

The top 5 countries by core-record count (United States, United Kingdom, Canada, India, Brazil) account for 454 of 1023 geolocated core records (44.4%).

n = 454 of 1,023 core records

Caveat: This corpus is a convenience sample of successful, publicly documented prosecutions and official findings — not a random or representative sample of religious-authority abuse. It cannot support prevalence, incidence, or per-capita claims about any tradition, country, or mechanism. Absence of a record is not evidence of absence of abuse; see research/ANALYTICS.md and static/data/COVERAGE.md. Geographic concentration in this corpus reflects, at minimum, English-language source availability and jurisdictional press/court-reporting infrastructure (see static/data/COVERAGE.md's Language coverage and Evidence-tier sections) as much as any true geographic concentration of the underlying behavior.

Data source: precomputed analytics pipeline. All figures computed at build time from 1,023 core records. See Method for full inclusion criteria.