Pillar weights — reasoning

Current weights (v0.1)

Pillar	Weight	Note
electoral	15 %	Electoral process and pluralism
governance	20 %	Separation of powers, legislative quality
judicial	20 %	Judicial independence, rule of law
media	15 %	Media freedom and plurality
civil	15 %	Civil liberties
corruption	15 %	Corruption and transparency

Sum = 100 %.

Status: v0.1 draft (2026-04-28). Requires review.

---

The weighting principle

Weights are a normative choice. This document explains why these particular values, not which value is objectively correct — no such value exists.

Considerations that drove the weights:

1. Empirical backsliding literature. Studies of democratic backsliding (Bermeo 2016, Levitsky & Ziblatt 2018, V-Dem reports 2020+) repeatedly identify attacks on judicial independence and on the quality of the legislative process as the most common and most effective channels of erosion. That is why judicial and governance get a slightly higher weight (20 % each).
2. Detectability between annual updates. The project’s goal is timely detection of directional change. electoral changes rarely (Chamber-of-Deputies elections every 4 years); media and corruption are tracked continuously. Higher event frequency in a pillar does not mean higher weight — that would distort the score. So frequency alone does not affect the weight.
3. Independence of dimensions. The pillars are designed to be as independent as possible (see pillars.md, section on overlap). Equal weights match the hypothesis that they are comparably important; the differentiated 15/20 reflects empirical evidence that some are channels of faster backsliding.

---

Discussion of alternatives

A) Equal weights (16.67 % each)

Pro: Simple, no subjective input, easy to defend as "neutral". V-Dem uses equal weights in some sub-indices.

Con: It implicitly says, for example, that full removal of judicial independence weighs the same as a local media incident of similar severity. Empirically this isn’t how democracies die.

Conclusion: Rejected, but equal weights are a useful control calculation — the index should also be displayed in this variant on the methodology page for transparency.

B) judicial and governance at 25 % each (the remaining 50 % distributed equally)

Pro: Stronger emphasis on structural dimensions, more consistent with the literature.

Con: Marginalises civil and media, which are very live in the Czech context (media concentration, minority protection). Risk that important trends in these areas disappear in the overall score.

Conclusion: Rejected as too aggressive. If validation shows that judicial + governance correlate with EIU/V-Dem more than other pillars, v0.2 could move towards 22.5/22.5.

C) Dynamic weights (context-dependent)

Pro: Adaptive to the situation.

Con: Destroys comparability over time. Backtesting stops making sense. Opens up vast room for motivated manipulation ("in this situation we’re weighting the judiciary higher because…").

Conclusion: Permanently rejected. Weights must be fixed within a version of the methodology.

---

Rules for changing weights

1. Changing weights = a new methodology version (v0.1 → v0.2). It’s not a fix, it’s a revision.
2. Every change requires:
   • A commit to weights.md and CHANGELOG.md with explicit reasoning.
   • Recomputation of the entire historical score series with the new weights and storage of both series (v0.x and v0.y) in data/scores/. The old version doesn’t disappear.
   • The change marked in the dashboard.
3. A change can only be proposed:
   • After quarterly validation (see methodology/validation_YYYY-Qx.md), if the index persistently and systematically diverges from EIU/V-Dem by > 10 points in a direction that cannot be explained by current events.
   • After a substantial methodology change at one of the primary sources (V-Dem fundamentally re-computes EDI, EIU changes sub-categories).
4. Weights can never be changed in response to a specific political outcome (election, crisis) — that would be a clear signal of motivated adjustment.