Ancap Welcomeness

0 means the weakest fit; 100 means the strongest fit with the model’s stated anarcho-capitalist preferences: voluntary exchange, property, self-defense, privacy, personal autonomy, monetary choice and the practical ability to live by those values. It is a transparent preference estimate, not an official country rating or the probability of a future event.

The original framework, with better evidence handling

Model ancap-welcomeness-v1.0.0 retains the original twelve structural pillars and their weights, and the original 60/40 balance. The 60% liberty component now uses admitted published indicators, qualified fiscal references and visibly downweighted original country priors. The 40% practical-living component uses safety, rights, basic services and a bounded contribution from the original accepted current-pressure model. This is a new model version, not a retrospective revision of old Bayes snapshots.

Liberty pillar	Weight inside 60%	Declared inputs
Self-defense	9%	Original country-specific self-defense prior; no invented global firearms-law coverage.
Property & contracts	12%	70% WGI rule of law, 30% WGI control of corruption.
Taxes & fees	10%	35% personal headline rate, 30% corporate, 25% consumption, 10% comparable mandatory entry cost; existing scope and review gates retained.
Business freedom	10%	WGI regulatory quality, explicitly a broad institutional proxy rather than a count of regulations.
Money & banking autonomy	10%	50% price stability, 50% original financial-autonomy prior. Inflation does not establish capital-control or self-custody law.
Privacy & expression	9%	50% V-Dem political civil liberties, 50% original privacy prior. The published index is an expression/association proxy, not a surveillance-law inventory.
Exit & mobility	8%	70% original exit prior, 30% unresolved residency permission. Citizenship and individual eligibility are not inferred.
Personal integrity & rule of law	10%	60% V-Dem physical integrity, 40% WGI political stability.
Family & education autonomy	6%	60% private civil liberties, 40% original family-autonomy prior; local schooling and family law remain gaps.
Independent infrastructure & access	6%	45% Internet access, 30% financial-account ownership, 25% original parallel-infrastructure prior. Access is not ownership or freedom from surveillance.
State size & fiscal load	6%	45% tax revenue/GDP, 35% central-government debt/GDP, 20% original state-size prior. Aggregate fiscal load is distinct from personal tax liability.
Jurisdictional choice	4%	70% original arbitrage prior, 30% unresolved subnational policy coverage.

The 40% practical-living component allocates 30% to homicide safety, 20% to physical-integrity rights, 10% each to electricity, drinking water and sanitation, and 20% to accepted current-pressure resilience. Current pressure therefore has a maximum 8% overall weight. Its original two-primary-source requirement and freshness gates remain; missing reports do not mean safety.

Explicit normalization and uncertainty

Published 0–100 WGI scores retain that scale. Published 0–1 liberties/rights indices are multiplied by 100. Access percentages retain their scale. Price stability is clamp(100 − 4 × |inflation − 2|); tax-revenue context is clamp(100 − 2 × revenue/GDP); debt context is clamp(100 − debt/GDP ÷ 2); homicide safety is 100 ÷ (1 + homicides-per-100,000 ÷ 5). These anchors and weights are authored choices, not fitted causal effects or validated Ancap thresholds. Rankings do not rescale themselves to force the current best country to 100.

Within every fixed-weight slot, the estimate is q × normalized value + (1 − q) × 50. Missing values use the explicitly neutral prior 50 and retain a full 0–100 sensitivity range. Their weights are never shifted to a favorable subset. Observations through two reporting years old retain full strength; ages three/four/five retain 80%/60%/40%, and older observations are excluded. Captures older than 14 days, future dates, conflicting duplicates and out-of-range values are excluded. Retained/overdue/error-marked sources within that window receive a further 70% factor. Retrieval health never changes the original reporting year.

Original authored priors require their original review date within 366 days and retain only 35% strength, with an authored ±15-point range before neutral shrinkage. They never count as published observations. Fiscal intervals retain the existing country/category/review/capture checks; the midpoint represents neutral missing components, not zero fees or a precise effective tax rate. A source’s published uncertainty endpoints and missing-input endpoints propagate through the fixed weights. The resulting range is source-bound and missing-input sensitivity, not a statistical confidence interval.

Coverage, ranking and provenance

Every directory country/territory has a numerical model estimate. A jurisdiction with no usable inputs shows 50, neutral placeholder, unranked. Ranked rows require at least 40% data-backed model weight; prior-led and limited-evidence entries remain searchable below them. This is an authored ranking threshold, not a reliability guarantee. The number is displayed alongside coverage, original-prior share, reporting years, strengths, restrictions and a complete input/weight/source export.

WGI and V-Dem components share measurement inputs. Repeated use has a fixed aggregate weight and never creates extra independent confirmations. Counts of source families are descriptive. Wealth, growth, inequality and other descriptive indicators remain available in the research tabs; they are not automatically converted into political or moral judgments. News keywords, absence of reporting, and CBDC project stages alone do not change the new headline score.

Loading, original model and replay

A separate no-network worker projects the existing public-source caches into directly served evidence files every five minutes. The opening score view reads latest annual observations and fiscal inputs together; histories and country tax details load on demand from the same public projection directory. The worker preserves original reporting, review and retrieval dates and never forces a provider refresh. This removes the busy analytics API from the country-score and annual/tax research loading path. The original broad threat scenario remains on demand. Observed readiness depends on device/network conditions; missing or expired evidence remains explicit.

Current estimates are paused during replay; archived scenario calculations and local history keep their original versions and formulas. After a live access check completes, an active workspace resumes its gated loading automatically; a response from an earlier owner or replay context is discarded.

The country directory is a reference inventory, not a sovereignty claim. National evidence does not settle subnational law, visa eligibility, tax residency, individual obligations or neighborhood conditions. Published evidence methodology · Tax source scope and admission rules.