Methodology

Reference document. UI labels are translated in Ranking.

Score version: v10 · Indicative data as of 2026-07.

This document explains how to read the ranking. It is not an official index:

liberty scores are structured judgements, not pure statistical measurements.

Liberty mode (7 dimensions)

Each territory receives a 0–10 score (no half points) on:

Indicative scale

• 9–10: among the world’s best on that axis
• 7–8: clearly favourable
• 4–6: median / mixed
• 1–3: clearly unfavourable
• 0–1: extreme (failed state, total control)

Overall score

\[

\text{score} = \sum_d \text{note}_d \times \frac{w_d}{\sum w}

\]

Weights \(w_d\) come from sliders or a persona. If weights do not sum to 100,

the score is renormalised (proportions matter, not the raw total).

Variants & map

• Several rows may share one country (Italy (Normal) / Italy (300k Flat), Spanish regions…).
• iso identifies the territory; map_iso (if set) picks the map polygon. Otherwise iso is used.
• Clicking a polygon shows every variant for that map_iso in the panel.

Expat regimes (toggle)

When enabled, the tax score for eligible countries is replaced by the

tax_score from the territory_regime toggle link into the regime catalogue

(Beckham, NHR/IFICI, lump-sum, FIG…).

Real-estate mode (focus, not exhaustive)

~30 freehold markets ranked + a few excluded markets listed separately.

Criteria: yield, rental/CG tax, demographics, debt, price trend, costs, credit, liquidity, landlord law, FX.

• Computed from raw inputs: yield, trend (overvaluation), and often capital gains (exempt years).
• Stored indicative scores: other s_* fields.
• Reference sources listed under the table (Numbeo, IMF, PwC, BIS…).

When capgains_*_exempt_years is set for the active holding mode,

the CG score comes from that duration (full exemption). Three distinct modes:

The RE ranking uses the selected mode (default: personal). If a mode

has no data, the score falls back to personal; the UI clearly shows

the gap. The Netherlands illustrates primary (eigen woning, 0%) vs

Box 3 investment.

Business mode (hub focus)

~40 jurisdictions. Corporate tax & CPI (corruption) anchored on public sources; the rest is indicative.

Weight profiles: Solo (exit + CIT + ease), Employer (labour + legal),

Holding (holding regime + exit + reputation + CFC).

CFC criterion: how light Controlled Foreign Company rules are **if you are

tax-resident in that jurisdiction** (high score = weak or no CFC).

e.g. FR 209 B / 123 bis, US GILTI/Subpart F, EU ATAD, vs Panama/Paraguay/UAE

often without notable CFC. Labels are indicative.

Tax risk & enforcement (s_enforcement): estimate of how severe

tax/legal controls on companies are (audits, substance, TP,

abuse of law…). High score = softer / more predictable climate (good for the

ranking). Distinct from reputation (blacklists / perceived compliance)

and CPI (corruption). Labels are indicative.

US LLCs (Wyoming / Delaware / New Mexico): same tax assumption: 0% as long

as there is no effectively connected US income (ECI). Ranking gaps

do not come from CIT, but from indicative scores:

Wyoming leads in Solo (ease / maintenance fees). Delaware can lead in

Holding on reputation. New Mexico stays behind on reputation.

Updates

Edit data/seed-data.json, run npm run seed (validates then reloads SQLite).

Reread this file when the grid or sources change (scores_version).

Decorrelation (v2)

The self, space and infra axes were adjusted to reduce correlation

with economy / property (city-states vs spacious territories, havens vs

public services, etc.). The goal is not zero correlation, but personas that

actually move the ranking.

Structured tax detail

When available, tax_detail also exposes income_min/max, inheritance_*,

corporate_* (numeric rates derived from labels) for comparison.

Homeschooling (self axis)

Indicative categorical status per territory (homeschool / SQL table), shown

in list detail and the map panel. It does not replace the self score (which

also aggregates conscription, health mandates, etc.).

Working source: data/homeschool.json (ISO + name overrides) merged

into seed-data.json at load. Not legal advice.

Jurisdiction types

jurisdiction_type ∈ country | subdivision | regime | freezone | microstate

with parent_iso / variant_of for map hierarchy.

v3: follow-on work

• Residualisation of property / money / tax vs economy (lower correlations).
• Variant badges on the map (centroid + ×N badge).
• Broader tax coverage (~170 jurisdictions) with min/max.
• parent_key + /api/jurisdictions endpoint.
• Runtime derivation of s_debt / s_rental from text when needed (lib/scoring.js).

v4

• Stronger residualisation of money / tax / self vs economy (pairwise > 0.70 removed except economy×property ~0.70).
• Indicative tax coverage ~220 / 238.
• Variant focus on the map: badge/map click recolours the polygon with the variant score; map-paint.js extracted.

v5

• Residualisation of property vs economy (pairwise > 0.55 except economy×infra).
• txcostScoreFromText + fallback in reFactor.
• GET /api/jurisdictions/:key + multi-variant hubs banner on the map.
• RE / Business comparators extracted into modes-re-biz.js.

v6

• Normalised SQL jurisdiction table (parent_key FK) populated at seed.
• SVG hatching on multi-variant polygons.
• GET /api/stats (coverage + correlations > 0.55).
• Docker probe requires the jurisdiction table.

v7

• Light residualisation of infra vs economy (no pair > 0.55 left).
• Min–max score ranges on hubs / tooltips / map panel.
• Richer CSV (parentKey, income, corporate).
• UI meta: tax coverage + hubs from /api/stats.

v8

• Indicative tax coverage 238/238.
• “Hide low confidence” filter.
• URL sharing: iso / tid for map focus; Escape closes the panel.

Tax confidence (review 2026-07)

~35 jurisdictions moved from low → medium/high when IR / CIT / inheritance

ranges are published (PwC WTS, Tax Foundation, national texts).

Opaque or unstable states stay low (see data/fiscal-confidence-audit.json).

Labels remain indicative ranges.

v9

• Property capital gains split into 3 holding modes: primary residence /

personal investment / via company (cg_ + s_cg_ scores).

• UI switch + URL hold= ; score and CG column follow the active mode.
• France / Netherlands (and other freehold markets) filled to distinguish regimes.
• Business: CFC criterion (tax residence).

v10

• Business: Tax risk & enforcement criterion (enforcement_text + s_enforcement).

High score = lower perceived severity. Weighted with CFC in Solo / Employer / Holding.

v11

• SQL tables regime (catalogue) + territory_regime (toggle links and dedicated rows);

GET /api/regimes; compat territory.regime for scoring.

• Map: public/regions.geojson (admin-1 polygons) + regional map_iso

(US-TX, ES-CT, CH-ZG…); panel / hubs grouped by family (familyIso /

parentKey).

Cost of operating a business (editorial overlay)

Outside the 0–10 score. File data/country-editorial.json

for all Business hubs in the seed (~43):

• cost_band: low | moderate | high | very_high
• setup_usd_min/max, annual_usd_min/max: indicative order-of-magnitude (USD)
• cost_note_fr / cost_note_en: setup + run reading
• sources[]: public URLs used to calibrate ranges
• hub_name: aligned on seed territory.name (join key)

Scope: formation, agent/secretary, address, lean accounting/audit.

Out of scope: founder salary, personal rent, personal income tax; share capital noted separately when material (DE/CH/LU…).

Always labelled indicative on country pages.