1. Identify the vintage
Every imported AMECO file stores its publication date, source URL and SHA-256 hash. SHA-256 is a 256-bit cryptographic fingerprint used to verify that a source file has not changed.
PUBLIC FINANCE · AMECO ·
AMECO UBLGAPS: building a genuine structural backtest without invented data
A reproducible method separating the estimate known when a budget decision was made from its later ex-post revision.
The problem: a structural balance is not a simple accounting snapshot
The European Commission’s AMECO database publishes UBLGAPS, the structural balance of general government as a percentage of potential GDP. The Spring 2026 release is dated 3 June 2026. Because potential output is estimated, values can be revised as macroeconomic data or methods change.
rule: a serious backtest stores two separate numbers: the estimate available when the decision was made and the latest revised estimate. The latter must never silently replace the former.
The four-step backtest protocol
2. Preserve the real-time estimate
For fiscal year t, the test should preferentially use the vintage that was actually available when that year’s budget was prepared or assessed.
3. Measure revisions separately
The difference between the latest estimate and the real-time estimate becomes a revision variable. It is not confused with the decision-maker’s fiscal performance.
4. Do not criminalise a statistical revision
Any personal liability must rest on legally defined conduct — falsification, concealment or deliberate circumvention — not on a later revision of estimated potential output.
What is already verified in the 2026 release
These checks come from DBnomics’ redistribution of the AMECO France series retrieved on 4 June 2026. does not fabricate missing annual observations: publication of the full series requires a direct, verified import from the source file.