How the test and matching work
How your answers become a position, how figures, parties and governments are estimated, and how the test keeps improving.
Scoring
Each answer ranges from −100 to +100. Every statement has a location: how extreme it is. Your score on each axis is the position (0–100) that best explains your pattern of answers, given those locations. Strongly rejecting an extreme statement says less than strongly rejecting a moderate one. The test also corrects for response style: it measures how far you typically move the slider and reads each answer relative to that, so cautious and emphatic answerers with the same views get the same score. Skipped answers are excluded, and coverage is displayed. Map 1 shows economics (60% ownership, 40% management) and social values (70% culture, 30% religion; views on taxes and inequality appear on the equality axis instead). Map 2 shows national identity horizontally and military intervention vertically. The authority scale under the maps is 50% civil liberties plus 50% governance. All three are drawn from the same nine axis scores, for you and for every figure, and a custom view plots any two axes directly. The response-style correction is applied conservatively: it is blended with “no correction” as if you had given ten extra answers at face value, so one answer cannot swing unrelated axes much on a short test.
Agreement percentages
Labels: 85%+ very strong match, 70–84% strong, 55–69% moderate, 40–54% weak, under 40% strongly opposed. Agreement compares your answers with theirs statement by statement, weighted by each question's weight: Agreement = 100 × [1 − Σ weight × |your answer − their answer| ÷ (200 × Σ weight)], over the questions you both answered. Identical answers give 100%, opposite extremes 0%, and a 2× question counts twice as much as a 1× question. Raw answers are compared as they are, so two people with identical answers always get 100%. Unanswered, skipped and historically inapplicable questions are left out entirely. At least 8 shared answers are needed. The same formula is used for figures, parties, governments, ideologies, election candidates and other users. Each comparison shows which of their answers are documented, taken from their record, or estimated.
Agreement versus similarity
Agreement compares answers. Ideological similarity compares positions: 100 × (1 − root-mean-square difference across the nine axes ÷ 100). Different answers can lead to similar positions, so the two numbers can differ.
How ideology positions are set
Each ideology's position is estimated from its core principles, then run through the test (estimated answers, then the same scoring as users and figures) so that ideologies, figures and you share one scale. The numerical percentages represent similarity to these models, not a probability, endorsement or verified correspondence with an ideological school.
Reference evidence
Political figures are scored from estimated answers to all 193 questions (see “How figures are placed” below), with every answer viewable on their profile. Government periods remain browse-only, or use a labelled ideology analogue, until documented policy evidence is reviewed.
Recommended research datasets
V-Dem: country-year governance indicators
How figures are placed
Each of the 181 figures takes the full 193-question test, using estimated answers. Their nine scores come from the same scoring formula as your own results, so figures and users share one scale.
1 · Estimate a stance on each axis
Each figure receives a 0–100 stance on each of the nine axes. The stance is based on their voting record, time in government, writings, platforms or broadcast output. A one-line basis is recorded for every figure.
2 · Generate an answer to every question
Each question has a location: the stance at which a respondent would answer neutrally. Moderate statements sit near 50. Extreme ones, such as "a religious authority should veto legislation", sit near the ends. The estimated answer is 100 × tanh(direction × (stance − location) / 25). A few cross-cutting statements, such as gun rights, drug policy, taxes, climate and school prayer, draw on two stances at once: gun rights, for example, tracks both civil liberty and cultural traditionalism. When the record shows a clear position on a specific statement, that answer replaces the generated one. Examples are Trump on tariffs and Glenn Greenwald on surveillance.
3 · Score with the test's own formula
The answers go through exactly the same location-aware scoring as user responses, including the response-style correction. They produce the nine axes, the two maps and the authority scale. Each question carries its own weight based on what it asks. Red-line statements count 2×: equal rights and eligibility for office regardless of religion or ancestry, freedom of minority worship, religious rule over civil law, free elections with opposition parties, freedom to criticize the government, and the defining statements of collective ownership and central planning. Other statements that define an axis outright and carry high stakes (one-party rule, abortion, collective ownership of industry, censorship, lockdowns) count 1.5×; concrete policies count 1.2×; ordinary items 1×; and vague or near-consensus statements, narrow technical policies, or questions largely driven by another axis count 0.7×.
4 · Validate
Average agreement between the odd-numbered and even-numbered questions on each axis (100 = identical).
Orderings that are not seriously disputed must hold by at least 5 points, such as Stalin above Sanders on autocracy.
Correlation with the original hand-placed positions of the 12 map figures (108 values). The mean absolute difference is 7.9 points.
Figures with a widely used ideology label (e.g. Trotsky → Trotskyism, Thatcher → Neoliberalism) should land near it. 68 of 82 rank that ideology in their 10 nearest of 130 (52 in the top 5). The ideology positions were set separately from these estimates, so this is the most independent check.
Average half-width of each axis's 90% range across 40 re-estimations. The re-estimations add stance error scaled to evidence depth (±5 high, ±9 medium, ±15 low) and ±20 error on every answer.
Estimates, not self-reports
Figures, parties and governments have not taken the test themselves. Their answers are estimated from their public records, and every estimated answer is shown on their profile so you can judge it and suggest corrections.
How the test improves
Every question has a feedback button, and the results page asks how accurate your result feels and lets you suggest corrections to any figure, party or government. That feedback is reviewed and used to reword confusing questions, fix unfair ones and correct placements.