PRIVI 2026›Methodology

Data imputation, normalization,
and weight determination

A systematic, reproducible pipeline for data processing and weighting, tailored to each indicator's distribution and cause of missingness.

01

Data Imputation

Step 1 · Temporal Backward Filling

Take the exact value for the target year where available; otherwise search back up to four years for the most recent observation.

Step 2 · Cross-Sectional Imputation

  1. Zero-value imputation — for count / aggregate indicators (7), where a gap means "not yet generated."
  2. Fixed-value imputation — for policy-absence indicators (3), such as the NDC3 ambition score.
  3. Global percentile imputation — for indicators with wide cross-country variation and sufficient coverage (26).
  4. Regional percentile imputation — for indicators strongly shaped by geography or development stage (11).
  5. Proxy indicator rank-transfer imputation — for systematically non-random missingness (2 indicators).

02

Indicator Normalization

Five normalization methods mapped to [0,1]

01

Min-Max

For positive indicators; linear mapping to [0,1].

02

Inverted Min-Max

For negative indicators, where larger values are worse.

03

Log-Min-Max

For highly right-skewed indicators spanning orders of magnitude.

04

Z-Score

For near-normal indicators with significant outliers.

05

Peaked

For indicators with an optimal range, modeled with a triangular tent function.

03

Weight Determination

Indicator weights (within group). Principal Component Analysis (PCA) within each sub-dimension group sets the third-level indicator weights, objectively reflecting each indicator's contribution to the group's common information and avoiding subjective weighting.

Dimension weights (across levels). The five first-level dimensions are aggregated with equal weights — the five core capitals are equally indispensable to long-term prosperity. The composite index runs from 0 to 1.

PRIVI is positioned as a complement, not a substitute. The assessments for North Korea and some Pacific island economies rely heavily on data imputation; individual ratings have limited precision and are best read qualitatively at the group level. This page is a public summary of the methodology — full formulas and parameters are in the source report.