How Caveat works
Caveat reads the same data as everyone else. The difference is what we do before publishing: check the denominator, check the basis, and print the objection a hostile reader would raise rather than burying it.
Every figure carries its source
Each number we publish exists as a row with its definition, unit, basis, publication date and a link to the series it came from. Every story has an evidence page listing those rows. If you cannot check a claim in about a minute, it does not run.
Sources are tiered
| Tier | What | Use |
|---|---|---|
| Tier 1 | Official statistical agencies: OECD, World Bank, IMF, Eurostat, ABS, RBA | Carries headlines |
| Tier 2 | Sector bodies, exchanges, central bank tables | Acceptable where no tier 1 series exists |
| Tier 3 | Commercial and crowd-sourced aggregators | Context only. Never carries a headline |
Nothing is estimated into existence
Missing values are never modelled, imputed or carried forward. Where a country lacks the data, it is reported as insufficient coverage rather than filled in. Figures we compute ourselves are marked derived, and the calculation is shown.
Indices are built to a published standard
Composite indices use fixed bounds set at creation, so a change in score over time means something rather than merely reflecting a change in the peer group. Equal weights are the default. Every index publishes its components, weights, coverage and a sensitivity test showing the ranking does not depend on our choices.
Published figures do not change quietly
Statistical agencies restate prior periods routinely. When that happens the revision lands in the next vintage with a changelog, not retrospectively into something already published. If a restatement moves a score materially, we say so.
We correct ourselves in public
A finding that contradicts something we published earlier is a story, not a threat. Corrections appear on the story they correct.
Caveat lector. Just saying.