The offshore corpora, counted

This page is our own analysis, not reporting. Every other page on this site points at an investigation somebody else published and follows what happened next. This one counts rows in a public database and shows the totals. It names no individual and makes no allegation, and it could not: the data cannot support one.

ICIJ publishes this database as a finding aid and states that inclusion in it is not evidence of wrongdoing — offshore entities have lawful uses. Nothing derived from it here is an allegation about any person or company.

Data: ICIJ Offshore Leaks Database (offshoreleaks.icij.org), Open Database License; contents CC BY-SA. Extract generated 2026-07-29.

What is actually in the database

⚠ These are counts of ENTITIES PUBLISHED AS STRUCTURED DATA, which is not the size of the leak. Pandora is the smallest row here and was among the largest leaks by document count — ICIJ published only part of it in this form. Read the table as what is available to analyse, never as which scandal was biggest.

⚠ sourceID in the raw files is per service provider, not per leak; Paradise and Pandora are each split across a dozen values. The rows above are grouped back into leaks.

Entities by jurisdiction

⚠ A jurisdiction's rank reflects which providers appear in these leaks, not which jurisdictions are most used worldwide. One provider's book is not the offshore industry. ⚠ Counts are grouped by jurisdiction NAME, because the same place carries several different country codes across these corpora — the British Virgin Islands alone appears under three. ⚠ 'Undetermined' is a value in the source data, not a residual we computed.

Incorporations by year — Panama Papers

⚠ 🚨 The tail of every series is an artefact, not a decline. Each corpus is current only to the date it was taken (the Panama Papers data stops in 2015), so the last years fall away because the leak ends, not because incorporations did. Never read the final points as a trend. ⚠ Entities with no incorporation date are excluded rather than guessed at.

Entities by service provider

What the records say became of the entities, by leak

⚠ Each corpus is current only to its own leak date, so these counts describe the provider's book as handed over — never what any authority did after publication. A low struck-off share is not evidence of inaction. ⚠ The five corpora came from different providers and registries; the columns are not filled in consistently, so compare within a leak, not across.

What this cannot tell you

• Whether anything was wrong. Inclusion is not evidence of wrongdoing, and no column in this data distinguishes a lawful structure from an unlawful one.
• What regulators did afterwards. Each corpus stops at its own leak date. Consequences arrived after that and are not in here — they are in the reporting, which is what the rest of this site follows.
• How the world changed. These are the books of the providers who were leaked, not the offshore industry. A fall in one provider's incorporations is a fact about that provider.