The Method
The CARO Fairness Index measures how fairly a population judges the arrangements it lives under. The instrument, the domains, the formulae and the versioning rules are published in full. Anyone may use them. CARO uses them on the same terms as everyone else.
Not satisfaction with a government. Not approval of a leader. The perceived fairness of the arrangements themselves — the structures that hold whoever is in office.
A respondent is asked to judge twenty-five specific governance practices on a five-point scale, from completely unfair to completely fair. The questions are about arrangements, not people: how a seat is vacated, how a public service is allocated, how an institution accounts for itself.
The instrument produces four things: a score for each question, a measure of how settled that judgment is, a score for each domain, and one figure for the whole. Each is defined below, and each is reproducible from the raw responses by anyone who has them.
Version One. Equal weight within and across domains; later versions may apply differential weights following review.
| Score range | Interpretation |
|---|---|
| 1.00 – 1.80 | Society strongly considers this unfair |
| 1.81 – 2.60 | Mostly unfair — significant concern |
| 2.61 – 3.40 | Mixed or contested — no settled consensus |
| 3.41 – 4.20 | Mostly fair — broad general acceptance |
| 4.21 – 5.00 | Society strongly considers this fair |
Complete and unabridged. Anyone holding the raw responses can reproduce every published figure from these alone.
Range 1.00–5.00. The raw judgment: how fair the respondents found this one practice.
Range 0.00–1.00. Near 1.0 means strong consensus; near 0.0 means maximum disagreement. This is what separates a settled judgment from a contested one — and a mean without it can describe nobody.
Range 0.00–1.00, capped at 1.00 regardless of overparticipation.
Prioritises the raw fairness judgment at 60 per cent, while rewarding consensus and broad participation at 20 per cent each.
Version One uses equal weights for all questions within a domain. Later versions may apply differential weights based on domain centrality assessment.
Version One uses equal weights across all five domains.
Range −1.0 to 1.0. Measures how far a specific policy moves with or against the fairness judgment the population has already expressed.
Updated continuously as new policies are assessed.
The method has been run in full on one cohort. These are the figures, with their limits stated.
Bangladesh, April 2026. All five domains scored below 2.00 — a broad consistent signal rather than concentrated unfairness in one area.
The cohort was 71 per cent aged 18–25, 62 per cent urban and 54 per cent undergraduate. It is not nationally representative, and no figure above should be read as a measurement of Bangladesh. It is a test of whether the instrument produces differentiated, interpretable results on a real population. It did.
Full results, every distribution and every standard deviation →
A measurement that changes without a published bridge destroys its own history.
The method is open. It is not CARO's property to license and not CARO's permission to grant.
Any person, organisation, researcher or government may apply this method to any population, publish the results, and describe them as Fairness Index measurements, provided the published method is followed and the provenance is stated. CARO claims no exclusivity over it and uses it on the same terms as any other person.
What CARO maintains is the standard itself: the definitions, the formulae, the versioning, and the record of what has been measured and by whom. The full specification is in Fairness as Foundation, chapter 13, listed with the rest of the published work on Publications.
Cite the version, not the organisation.
Where a figure produced by this method is cited, carry its provenance: what was measured, where, when, how the sample was drawn, and what was excluded.