pre-registration-first economic framework

Economic politics should be settled by testable claims.

IESET is a public evidence map for economic policy. Instead of asking which tribe sounds most confident, it asks every school to make claims that can lose: what should happen, where, over what period, and what data would change our mind. Those claims are registered before analysis and scored against reproducible runs.

Start with a claim, a school of thought, a policy, or a country. The library links each one through shared policy levers so a slogan can be traced down to its test, data vintage, code, steelman, and scoreboard impact.

The scoreboard shows raw forecast scores alongside stricter evidence-quality and attribution gates. When schools remain inside the strict no-call band, the honest conclusion is that the current corpus has not separated them at high integrity — not that any school has been vindicated. See the scoreboard.

Census (2026-07-18): 1833 hypotheses · 1818 runs · 3358 policies · 700 movements · 17 positions · 4,938 school predictions · 1,835 result cards

Search hypotheses →Open the scoreboard How it is produced How it works

How a hypothesis becomes a verdict

The loop is locked down so arguments have to leave a paper trail. The pass/fail rule is written first, then the data is fetched, then the rule runs against the data. Nothing about the rule moves once the data is touched — that's the whole point.

Write the rule first

The hypothesis YAML names the metrics, the threshold, and the estimator. It's committed to git before any data is fetched. The commit timestamp is the proof the rule didn't bend to the result.

Pull from primary publishers

Series are fetched from the original sources — World Bank, IMF, OECD, BIS, FRED, Eurostat — frozen as a versioned vintage with a content hash. Old vintages are never overwritten, so every run stays reproducible from the bytes it actually saw.

Run the locked rule

The estimator runs against the pinned vintage. The pre-registered threshold mechanically decides supported, partial, or refuted. Every result ships with the steelman — the strongest argument against what we found.

Data comes from public publishers

World BankIMFOECDBISFREDEurostatBLSECBBoEMaddisonShillerV-DemPolity5Fraser EFWHeritageILOWHOUN DESAFAOIEAWIPOUN ComtradeSIPRIFreedom House… and more in the source registry

Read the full methodology →Browse all hypotheses

Axes — the framework's spine

Party labels are a bad measurement device. "Left" and "right" mean different things in different countries, and campaign mood is not evidence. The framework codes every policy by channel — fiscal, regulatory, monetary, institutional — so the same content is scored the same way no matter who enacted it.

Each policy moves axes

A tax-cut policy moves fiscal.tax_progressivity in the − direction.

Movements aggregate

A government's axes_summary fingerprints what it actually did, regardless of how it labelled itself.

Hypotheses test

When axis X moves direction Y, does outcome Z follow? Schools that predicted right or wrong are scored on the Scoreboard.

Browse the axes →Why this design

Recent results

Open the hypothesis library →

verdict keysupported— data ran as predicted, cleared thresholdpartial / mixed— some predictions held, others didn'trefuted / weakened— data ran opposite or failed thresholdrun pending— pre-registered, model not yet fired

supportedmonetary