How the police index works

Calculation

All pillars run from 0 to 100. WJP inputs below are the published fractions from 0 to 1.

T = published Eurobarometer police trust percentage
I = 50 × (lawfulness + 1 − widespread corruption)
P = 50 × (respect for basic rights + avoidance of excessive force)
A = 50 × (complaints investigated + lawbreakers punished)
R = 100 × V-Dem v16 v2x_clphy, year 2025

Index = 0.20T + 0.20I + 0.20P + 0.20A + 0.20R
Perception-only comparison = (T + I + P + A) / 4

Each pillar contributes up to 20 points. Integrity, treatment and perceived accountability each contain two equally weighted survey items. No country-size adjustment or EU min–max stretching is applied. The rights component does not cover all unlawful police killings, routine excessive force or complaint-handling outcomes. In particular, small differences between countries’ generally high rights-context values are not expanded to fill the entire 0–100 range.

Finland’s unrounded example:

Trust                    93.00000000 × 0.20 = 18.60000000
Integrity                90.76066955 × 0.20 = 18.15213391
Treatment                74.20634800 × 0.20 = 14.84126960
Perceived accountability 78.15741635 × 0.20 = 15.63148327
Rights context           98.70000000 × 0.20 = 19.74000000
Total                                        86.96488678
Published points                                    87.0

The weights were fixed before calculating the revised ranking. The complete scored dataset has 216 source observations: seven police survey inputs and one V-Dem index per country. Source precision is kept for calculation; it does not imply equivalent measurement accuracy.

Rules and limitations

Units and names

One national-system entry per EU27 country. Named services identify the policing system, but the source surveys do not assign scores to individual forces or establish which forces respondents had in mind. The directory uses official force, government and local-tier websites.

Scored scope

ECHR case counts, CPT ordinal codes, oversight codes and transparency codes receive no points in this edition. Incomplete research is preserved separately. All 27 countries use the same complete five-pillar design; there is no country-specific redistribution of missing weights.

Weights

Trust 20%; integrity 20%; treatment 20%; perceived accountability 20%; rights context 20%. Fixed before calculating the revised ranking. Each scored pillar is required. Weights are editorial choices, not fitted estimates.

Trust

Published percent tending to trust police, Standard Eurobarometer 105, QA6.4. Fieldwork 12 March–5 April 2026. Keep the published response base including don’t know. Source

Integrity

50 × (lawfulness fraction + 1 − widespread-corruption fraction). Both WJP fractions are published national estimates. Reverse corruption includes don’t know and is not the percentage of honest police. Source

Treatment

50 × (fraction agreeing basic rights are respected + fraction agreeing police avoid excessive force). WJP EUROVOICES, national Total Sample estimates. Source

Perceived accountability

50 × (fraction agreeing complaints are investigated + fraction agreeing police lawbreakers are punished). Public belief, not actual complaint outcomes or sanctions. Source

Survey dates and samples

WJP country fieldwork spans November 2023–May 2024. The total 64,089 interviews is not the item-level sample: questions were assigned to random questionnaire groups. Sources differ in ages and interviewing modes. Do not infer statistical significance from small rank gaps. Source

Rights context

100 × V-Dem v16 v2x_clphy for 2025, the latest common year for all EU27. The source combines freedom from government torture and political killings. Includes government agents beyond police. Higher is better; do not invert or rescale within the EU. Source

Rights scope and influence

Rights context serves as a broad context check and barely separates EU states: its contribution ranges from 17.40 to 19.78 points, a cross-country spread of 2.38 points. Relative to the perception-only league, nine countries change rank, by at most two places. The 2.38-point spread is not a bound on score changes when the fifth pillar is removed. The measure covers government agents broadly and cannot establish any force’s record.

Rights uncertainty

Retain source codelow and codehigh. Rights-only composite bounds hold survey estimates and weights fixed and vary only this component. These are not confidence intervals for the full composite. Expert assessments can reflect reporting and model limitations. Source

Rank and display

Competition ranks use unrounded totals. Display points to one decimal. Exact score ties share a rank; alphabetical order breaks display ties only. Identical displayed points can have different ranks.

Score bands

Top ≥85; High ≥75; Moderate ≥60; Low ≥45; Very low <45. These are editorial score bands, applied to unrounded totals. They do not establish a probability of harm.

85 weight scenarios

Baseline + each of 5 weights varied by ±5,10,15,20,25,30,35,40% relative, with the other four rebalanced proportionally: 81 vectors. Add rights weights 0,5,10,30%, with the four perception weights equal: 85 unique vectors. These are deterministic alternatives, not probabilities.

Sensitivity and custom weights

In the calculation workbook, changing active weights updates the main league. The 85 published scenario vectors stay fixed; source input changes still update their scores and ranks. Its opening note identifies custom weights. The published web table retains the fixed baseline and its perception-only comparison. A separate check removes reverse corruption and uses lawfulness alone for integrity.

Comparison across versions

This edition makes no year-on-year comparison. Differences between alternative draft methods or the perception-only comparison reflect the calculation, not measured improvements in policing. Development version numbers are not a historical time series.

Research context

Selected external oversight and review mechanisms are described with sources in every country profile, without scores. Parliamentary bodies, ministerial investigators and custody monitors have different remits. Historical CPT observations and uploaded codes remain research records, not verified outcome scores. Missing reports never establish a clean record.

Source and edition

V-Dem Country-Year Dataset v16, released March 2026; observations for 2025. The exact official GitHub source used for this edition is pinned and hash-verified. Do not mix V-Dem release versions for longitudinal comparisons. Source

Attribution and reuse

Original composite by LeagueIndex, without endorsement by source organisations. V-Dem’s current FAQ and dataset page confirm CC BY-SA 4.0 and request both the dataset and measurement-model citations. Derived scores and authored methodology are supplied under CC BY-SA 4.0; original third-party materials retain their terms. Checked 18 September 2026. Source Source

Principal force selection

The directory identifies principal national and constitutionally or statutorily distinct regional services whose remit includes general public-facing policing, public order or border policing. It includes gendarmeries and separate border services. Regional branches of a national service are not separate entries. Specialist criminal-investigation-only, financial, customs, intelligence, prison, parliamentary, military-only and maritime-only services are excluded. Ordinary municipal services are not enumerated. Belgium’s statutory local policing tier is explicitly identified, with its official zone directory; it is not presented as one force. Germany’s territorial services cover all 16 Länder, including the separate services of Bremen and Bremerhaven. This is a directory of organisations and a local tier, not a separate league or an exhaustive list of every policing unit.

Minimum data

All five valid pillars are required. Missing values never become zero or disappear through redistributed weights. An observed zero remains valid. The rule is also applied to the published sensitivity runs, so they do not conceal missing source data. The present edition has no missing scored cells.

Weight sensitivity results

There are 85 unique published weighting alternatives, producing 2,295 country–scenario results. Finland is first in every scenario. The widest rank span is four places. The scenarios explore editorial choices; they do not quantify survey sampling error or the probability of a country’s rank.

The source’s rights-context bounds are carried into a separate partial uncertainty calculation. These bounds vary the rights component alone. The survey components also have uncertainty, and the package does not provide a valid overall statistical confidence interval. An additional indicator check replaces the reverse-corruption component with lawfulness alone, addressing its sensitivity to “don’t know” responses.

Source citations

Coppedge, Michael, et al. 2026. V-Dem Country-Year Dataset v16. Varieties of Democracy (V-Dem) Project. DOI: 10.23696/vdemds26. Dataset DOI. The extract is pinned to the official repository commit f4dd26922e658442524dfd954bf14f7ebe622d5d.

Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2026. “The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data.” V-Dem Working Paper No. 21, 11th edition. University of Gothenburg: Varieties of Democracy Institute. V-Dem’s current citation guidance. Both citations are requested for v16. The guidance and licence were checked on 18 September 2026.

Standard Eurobarometer 105; WJP EUROVOICES data and codebook; WJP questionnaire; WJP methods; V-Dem codebook.

Reuse

The V-Dem extract, derived scores, and authored methodology are supplied under CC BY-SA 4.0, with attribution to V-Dem and LeagueIndex as applicable. The original Eurobarometer, WJP and other source materials retain their own terms. Keep the metric definitions, source years and scope statement with published tables.

V-Dem confirms its data licence on the dataset page and in its FAQ. Country directory and oversight source links are provided alongside the relevant entries.