Research governance
Correcting a published result.
A composite assessment of 195 governments will contain errors. The question that determines whether it can be cited is not whether they occur, but whether they are found, published and carried through to the score.
What can be disputed.
Any indicator value, any source attribution, and any statement of fact in a published report. Disagreement with the framework itself — the choice of indicators, or the decision to weight the six pillars equally — is not a correction. It is a methodological objection, and it is answered in the methodology rather than by changing a score.
A dispute is strongest when it identifies the indicator, the value recorded, the value contended, and the public evidence supporting it. Evidence that is not publicly verifiable cannot enter the assessment, because a result no reader can reconstruct is not a published result.
The sequence
Receipt is acknowledged within five working days. The indicator owner reviews the evidence against the handbook definition and responds with one of three outcomes: the value is corrected, the value is confirmed with the reasoning stated, or the definition itself is found to be ambiguous and is scheduled for revision in the following edition.
Where a correction changes an overall score, the ranking is recomputed and republished. Where it changes only an indicator, the indicator dataset is reissued with a new version number. Neither is done silently: the previous value remains visible in the correction log.
What is recorded
Every accepted correction is logged with the date it was received, the date it was applied, the indicator and government affected, the previous and revised values, and the effect on the overall score and rank. The log is public and is not pruned between editions.
Rejected disputes are logged in aggregate — how many were received and how many were declined — without identifying the party that raised them, so that the volume of challenge is visible without discouraging it.
What is never done
A score is not adjusted in response to the standing of the party requesting it, nor withdrawn because a government objects to its position. Corrections are applied on evidence, and a government that supplies better evidence than its neighbor will see its indicator corrected on that basis alone.
The rest of the policy
Right of reply, versioning and escalation.
-
Who may raise one
Anyone. A correction is assessed on the evidence attached to it, and the standing of the party raising it is not a factor in the outcome. A government disputing its own indicator, a researcher who spotted an arithmetic error and a reader who followed a broken source link are all handled through the same queue.
-
Right of reply
An entity named in a published result may submit a response for publication. Where the response is substantive and evidenced, it is published alongside the result — whether or not the figure changes — so that a reader reaching the score also reaches the objection to it. A response is not edited for position, only for length and for material that would be unlawful to publish.
-
How quickly
Receipt within five working days. A plain factual error — a transposed value, a broken citation, a mislabelled year — is fixed as soon as it is confirmed. A dispute requiring judgement against an indicator definition is answered within thirty days, and where that is not achievable the delay and its reason are given rather than left to silence.
A defect that materially misstates a published rank is corrected out of cycle rather than held for the next edition.
-
How a change is versioned
Corrections are applied by issuing a new version, not by editing a file in place. Each dataset release carries a version number and a date; the correction log records which version introduced a change and which one it superseded. A copy taken before the change can therefore be identified as superseded rather than merely disagreeing with the current figure.
-
Withdrawal
Where an error is severe enough that a result cannot be repaired — a source that turns out to be invalid, or a method fault affecting a whole register — the result is withdrawn rather than quietly adjusted. A withdrawal notice stays published in place of the figure, states what was wrong, and remains there after the corrected result appears.
-
If you disagree with the outcome
Reply and ask for the decision to be reviewed. The review is carried out by someone who did not make the original determination, and the reasoning is given in writing. Where the disagreement is with the framework rather than the value, it is recorded as a methodological objection and answered in the next methodology revision — visibly, so that the objection is on the record even when the score does not move.
-
Where to send it
Write to [email protected] with the subject line beginning Correction. Include the register and entity, the indicator, the value published, the value contended, and a publicly accessible source for it. Evidence that cannot be verified by a reader cannot enter the assessment, however credible its author.
The correction log referred to throughout is published once the first edition carries assessed figures. Until then, this policy states the commitment; there is nothing yet to log.
Discuss the 2026 findings with the research team.
Methodology questions, indicator-level data requests, briefings for policy and executive audiences, and media inquiries are handled directly by the assessment team.