0x

Goal: test whether very high wealth is linked to higher documented Epstein-network contact in public records. This page tests a causal hypothesis in stages. It does not claim legal guilt for any person.

As of February 10, 2026 ยท USSC FY2024 / Forbes / Census / DOJ public releases

STEP 1: VISUAL CHECK

Model curve and observed bins,

displayed side by side.

Filled bars are the normalized model estimate. Outlined bars are what we directly see in the sample. We show both because causality depends on whether the pattern survives outside model assumptions.

WHAT THIS SUPPORTS

After normalization, the model shows a mild positive slope across wealth levels.

WHAT THIS DOES NOT PROVE

A final causal effect from observed data alone, or any individual legal conclusion.

M:0.2%
O:100% (n=4)$1-$2B
M:0.4%
O:80% (n=5)$2-$5B
M:0.9%
O:75% (n=4)$5-$10B
M:1.8%
O:67% (n=3)$10-$25B
M:4.5%
O:N/A (n=0)$25-$50B
M:8.8%
O:N/A (n=0)$50-$100B
M:16%
O:100% (n=3)$100-$200B
M:28%
O:0% (n=1)$200B+

Filled bar = model estimate (M). Outlined bar = observed sample rate (O).

$10B model = 1.8%
$1B$500B

At $10B, the model-implied probability is 1.8% under current parameter choices. This value is scenario-dependent.

STEP 2: CAUSAL STRESS TESTS (n = 20)

Dose-response hypothesis test

Signal is weak in observed data.

Trend test across wealth bins (Cochran-Armitage)p = 0.3173
Main small-sample model (Firth logit)OR 0.778 [0.443, 1.368], p = 0.3835
Shuffle-label check (exact permutation)p = 0.3827
Resample stability (bootstrap, 5,000)OR CI [0.000, 1.685]
One-row sensitivity (leave-one-out)OR 0.638 to 1.047
Mislabel sensitivity (flip 1/2/3 rows)OR>1 in 5.0% / 15.8% / 24.3%

Plain-English read: once we remove strong model assumptions and test directly on observed rows, evidence for a positive causal slope is mild to weak.

n = 20 coded entities.

Public records. Statistical sample.

20
ANONYMIZED ENTRIES
16
POST-CONVICTION FLAGS
Day 769
SINCE RELEASE DATE

Dataset definition: anonymized entities that met this project's public-record coding rules. This section reports associations in the sample and does not assert legal liability or guilt.

Billionaire A$10B-$25B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire B$5B-$10B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
Billionaire C$100B+
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire D$2B-$5B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire E$5B-$10B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire F$1B-$2B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire G$2B-$5B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire H$100B+
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire I$10B-$25B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire J$100B+
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
Billionaire K$100B+
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire L$1B-$2B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire M$5B-$10B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire N$2B-$5B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
Billionaire O$5B-$10B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire P$1B-$2B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire Q$5B-$10B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
Billionaire R$1B-$2B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire S$2B-$5B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION
Billionaire T$2B-$5B
PUBLIC RESPONSE: NONE LOGGED
Dataset entry coded from public records under this project's criteria.
POST-CONVICTION

If new primary-source records or direct public statements emerge, this page should be updated with a dated correction note.

STEP 3: SHOW THE WORK

Full data. Full code. Full assumptions.

ANONYMIZED SAMPLE: WEALTH vs MODEL-IMPLIED PROBABILITY

0%10%20%30%40%50%60%70%$1B$2B$5B$10B$25B$50B$100B$200B$400BNet Worth (log scale)
Finance
Tech
Real Estate
Media
Other
Dot size = document mentions

Each dot is an anonymized dataset entry. Dot size reflects mention count in released documents. The red line is model-implied, not an observed fit line.

For row-level roster details, use the primary anonymized list section. This tab focuses on the geometric relationship and model overlay.

The data shows an observational dose-response association in this dataset. Use it as a hypothesis-generating signal and pair it with transparent sourcing and cautious interpretation.

All claims sourced from public court filings, government databases, DOJ releases, sworn depositions, flight logs in evidence, or individuals' own public statements. Appearance in documents does not imply guilt. No individual named has been charged in connection with Jeffrey Epstein. This page is for exploratory statistical reporting. Any named individual who wishes to provide a statement or correction may contact us and it will be published in full. For publication risk, obtain media-law review.