Methodology
A full description of how the index is built — from the structural baseline through weekly event classification to the oversight model that holds quality without mandatory pre-merge review. Every document is live, versioned in Git, and changeable only through commits with reasoning recorded in the CHANGELOG.
Main documents
The six pillars
What each of the 6 pillars (elections, governance, judiciary, media, civil rights, corruption) measures, how it maps onto sources, and what doesn’t belong in it.
Read →Severity rubric
Five-point severity scale 1–5 for events, with concrete Czech examples, escalation/de-escalation rules, and "needs_review" criteria.
Read →Pillar weights
Reasoning for the current 15/20/20/15/15/15 weights, discussion of alternatives, and rules for changing weights in the future.
Read →Oversight model
Six layers of oversight (self-audit, source-count cap, daily reports, anomaly detection, monthly spot-check, public dispute) instead of mandatory pre-merge review.
Read →Structural mapping
How exactly the structural baseline is computed for each pillar from V-Dem 2024 / EIU 2024 / FH 2025 / RSF / TI / WJP.
Read →Data sources
Where the index pulls from — 8 Czech newsrooms, open data from parliament and the courts, watchdog organisations, international news. Live table generated from config/sources.yaml.
Read →Public opinion
Supplementary read-only context from polls (CVVM, STEM, Median). Doesn’t feed into the score — why not, how to use it, what we plan to add. Sources and their profile.
Read →Country comparison
How the 8 countries (V4 + DE/AT + USA/UK) and 6 indices were chosen, why CZ + SK are highlighted, and which publication years are used. A read-only external benchmark — does not feed into our index.
Read →Audit trail
Quarterly validation reports
Auto-generated comparison of our index with external benchmarks (V-Dem, EIU, FH, RSF, TI CPI, WJP). A persistent divergence > 10 pts across two quarters triggers a methodology review.