How every number was made
Credibility is the product. This page says what each figure on the site is, what it is not, which window each page uses, and how to check it yourself.
What the figures are
Three kinds of number appear on the site, and each is labeled with its kind in the sentence it sits in, not in a footnote.
- Among quiz takers
- Completions of the World's Smallest Political Quiz in the stated period: completed attempts, live records, bot and staff attempts removed, scored on the quiz's standard grid. The unit is a completion, not a person. A booth tablet or a classroom of shared laptops produces many completions from one device, and those are real people, so repeat completions are kept; the bot filter removes abusive ones. Counting one completion per device instead would move the libertarian share by under a point. They chose to take a political quiz; they are not a sample of anyone else. This is what the home page, the chart and the state pages show.
- Among Americans
- Estimates from a national survey of American adults commissioned by Advocates for Self-Government, fielded in 2026. This is the report's benchmark, printed beside the quiz figure and never in place of it. Its method is published with its results.
- Since 1994
- Advocates for Self-Government's cumulative count of completions across every version and platform of the quiz since 1994. The current database holds the attempts since 2011; earlier years come from the organization's records. The figure on the home page is that count, rounded down.
- Voted Libertarian
- The Libertarian Party's share of the certified presidential vote, from the Federal Election Commission, by state and nationally. The comparison every page makes.
The instrument
Ten statements, two scores, one hundred and twenty-one places to land.
Five statements measure personal freedom and five measure economic freedom. Each is answered Agree, Maybe or Disagree, worth 20, 10 or 0 points, so each axis runs from 0 to 100 in steps of ten. Agree always means less government, so no statement is scored backward. The pair of scores is looked up in a fixed 121-cell matrix that assigns one of five types: libertarian, conservative, progressive, authoritarian, moderate. The matrix is published; the site's chart draws it cell by cell.
The quiz also asks, before the result is shown, what the person calls themselves. The gap between that answer and the scored type is what this site calls the politically homeless.
| Version | In use | On this site |
|---|---|---|
| 5.0.0 | to November 2024 | The state pages: January to October 2024, the last window with ZIP codes. |
| 6.3.0 | from June 2026 | Current. Everything else on the site, from July 1, 2026. Four of the ten statements differ from 5.0.0. |
Which window each page uses
Two windows, never joined into one line.
From July 1, 2026
The home page, the chart, the definition page and the report: everyone who completed the quiz on the current version, from its first full month, refreshed as new completions arrive. The window is printed in every footer.
January to October 2024
The state pages: the last window in which the quiz recorded a ZIP code, so the last that can be counted by state. It ran on version 5.0.0, which shared six of the ten statements with the current version, so its national share (19%) and the current one (29%) are different instruments as much as different people. The two are never shown as a trend.
When ZIP capture resumes, the state pages move to the current version and carry both windows, labeled.
The survey beside it
Two instruments, related and not identical.
The quiz measures the people who take it. To measure the country, Advocates for Self-Government commissioned a national survey of American adults, designed with and fielded by a research vendor in 2026. It asks about the same two kinds of freedom, personal and economic, and it asks people what they call themselves, so its results can be read beside the quiz's.
The two instruments differ in wording and in how a score is turned into a type, so the site never converts one into the other. When the survey reports, this section will carry its sample design, its classification rule, and the crosswalk to the quiz, with the limits of that crosswalk stated.
Until then
Every figure on this site is a quiz figure and is labeled as one. The survey's figures will appear only on the report and here, labeled as survey figures.
The five phrases
The site uses plain words for the groups the quiz measures. The research terms are listed here once, and nowhere else.
| On this site | Defined as | Research term | This window |
|---|---|---|---|
| hold libertarian positions | scored in the libertarian region of the matrix | Inference | 29.1% |
| call themselves libertarian | answered "libertarian" to the self-identification question before seeing a result | Identity | 5.8% |
| both | the two above, together | Intersection | 3.9% |
| the positions without the word | scored libertarian, answered anything else | Positions without the label | 25.2% |
| the word without the positions | answered "libertarian," scored elsewhere | Label without the positions | 1.8% |
Shares are among the 51,238 people in the window who answered the self-identification question; 20,647 did not and are counted in no group. The segments block in current.json carries these counts every month.
The state score
One number per state, 0 to 100, from five parts. Every part is published by someone else or shown in full here, and every state page prints its five marks beside the weights.
Each part becomes a mark from 0 to 100, where 100 is the freest or most libertarian state on that part. The score is the weighted average of the marks. A state missing any part is not scored; that is only the District of Columbia, which Cato does not rank.
The quiz part needs two adjustments, and both are conservative. First, small states have wide margins (Vermont's share rests on 388 quiz takers), so each state's share is pulled toward the national share as if 1,000 average quiz takers were added to it. Second, because the shares run only from about 15% to 22%, the adjusted share is ranked against the other states rather than used raw, so the part spreads the same 0 to 100 as the others. The vote part is ranked the same way.
What the score is not: a measure of residents. The quiz part describes people who chose to take the quiz with a ZIP code in the state. The other four parts describe the state's laws and its voters. Weights are an editorial choice, stated here so they can be argued with. The ledger of wins is not in the score, because it records what we have verified, not everything that happened.
| Part | Source | Turned into a mark by | Weight |
|---|---|---|---|
| Personal freedom | Cato, Freedom in the 50 States, 2023 rank | rank 1 of 50 = 100, rank 50 = 0 | 25 |
| Economic freedom | Cato, Freedom in the 50 States, 2023 rank | same | 25 |
| Civil forfeiture law | Institute for Justice, Policing for Profit, 2025 grade | A 100, B 75, C 50, D 25, F 0 | 10 |
| Quiz takers libertarian | The quiz, January to October 2024, ZIP mapped to state | share pulled toward the national share by sample size, then percentile among states | 25 |
| Voted Libertarian | FEC certified 2020 presidential results | percentile among states | 15 |
Publication rules
What keeps the site honest, and inside its charter.
- Aggregates only.
No individual record reaches this site. The quiz data arrives as counts by cell, type and self-identification. No survey respondent's record reaches this site either.
- State is the finest geography.
Nothing below the state is published. Any cell under one hundred people is suppressed.
- Every figure says what it covers.
The period, who is counted, which version of the quiz, and what was filtered out, in the sentence and not in a footnote.
- Two instruments are never merged.
Quiz figures and survey figures appear side by side, each labeled. Neither is converted into the other.
- No trend across an instrument change.
Where the statements changed between versions, the windows are shown separately and never joined into one line.
- The site never asks a scored question.
The quiz lives on theadvocates.org. This site shows; the quiz measures.
- No candidates, no endorsements, no mobilization.
The state pages list organizations doing the work; they never list who to vote for.
Downloads and citation
Everything on the site can be checked from these files.