The method

What “left” and “right” mean here

The words come from the 1789 French National Assembly (reformers sat left of the chair). In American politics they now bundle two things that do not always travel together: how much the state should steer the economy, and how much it should police culture, borders, and bodies. Spectrum’s composite is the American bundle. The compass splits it.

The two academic engines we borrow

We do not scrape a live feed of every roll call. We do use the logic of the two methods that actually rank Congress, and we blend published scores into the composite when a sitting member has one.

DW-NOMINATE — votes

Poole and Rosenthal’s DW-NOMINATE (Voteview) places legislators in a space where distance means “they vote differently.” The first dimension, running about −1 (liberal) to +1 (conservative), is the familiar economic/governmental left–right line and is comparable across Congresses. A second dimension sometimes captures intra-party fights (civil rights, nativism, lifestyle). Pew, political scientists, and journalists use the first dimension as the standard academic spectrum. It only exists for people who cast roll-call votes. Presidents, governors, and first-time candidates have to be estimated some other way.

GovTrack ideology — cosponsorship

GovTrack builds a who-cosponsors-whom matrix for the House and Senate separately, looking at the current Congress plus the previous two (roughly four to six years). It then runs principal components via SVD. The second dimension of that decomposition tracks left–right in practice, even though the procedure is blind to party and to bill text: people who cosponsor the same authors sit together. Leadership is a separate PageRank on the same matrix. Cosponsorship is cheaper than a tough vote, so it can be gamed; it is still one of the cleanest behavioral signals besides roll calls.

Spectrum’s composite

Each figure starts from an ideological archetype (democratic socialist through MAGA House), then takes issue-level offsets from the public record. Twelve issues are scored from −100 (left pole) to +100 (right pole). If a legislative first-dimension score exists, it is mapped onto the same −100…100 scale and blended:

overall = 0.42 × (NOMINATE × 100) + 0.58 × issueAverage
issueAverage = Σ (issueScore × weight) / Σ weights
If there is no roll-call history, overall = issueAverage.

That 42/58 split is a judgment. Votes are the hardest evidence and the narrowest: they miss governors, miss candidates, and miss the person who votes with leadership on must-pass bills but campaigns as a firebrand. Issues restore that. They also let a libertarian (Massie, Paul) sit “far right” on the line while dropping down the compass.

The twelve issues

Weights sum to 1. Economic issues pull the composite toward taxes, health, climate, labor, and regulation. Social issues pull toward immigration, abortion, guns, crime, schools, and civil liberties. Defense sits in both.

• Taxes & spending11%

  Left: Higher taxes on high earners, larger public budget. Right: Lower taxes, smaller fiscal footprint.

• Health care9%

  Left: Public or single-payer coverage as a right. Right: Private insurance markets, fewer mandates.

• Climate & energy8%

  Left: Aggressive climate rules and fossil-fuel phase-down. Right: Energy abundance, fewer climate mandates.

• Business regulation9%

  Left: Stronger antitrust, labor, and consumer rules. Right: Deregulation and lighter administrative state.

• Labor & wages7%

  Left: Unions, higher minimum wage, worker power. Right: Flexible labor markets, right-to-work.

• Immigration10%

  Left: A broader legal path, less interior enforcement. Right: Tighter borders, more deportations.

• Abortion9%

  Left: Legal access, opposed to bans. Right: Restrictions or bans, fetal personhood.

• Guns8%

  Left: Licensing, assault-weapon limits, red-flag laws. Right: Broad Second Amendment, fewer federal limits.

• Crime & policing7%

  Left: Reform, diversion, less incarceration. Right: More policing, longer sentences, less bail reform.

• Schools & speech6%

  Left: Public-school investment, DEI frameworks. Right: Choice, against gender/DEI curricula.

• Defense & alliances8%

  Left: Diplomacy-first, smaller overseas footprint. Right: Hard power, high defense spending, deterrence.

• Civil liberties8%

  Left: Privacy, speech, limits on surveillance and bans. Right: Order, platform / campus / speech restrictions for security or morality.

Place yourself uses the same twelve issues and the same composite. The full path is thirty agree/disagree items (two or three per issue, with reverse-scored stems so “always agree” does not shove you to one pole). Express is twelve items, one per issue. Likert answers are mapped to −100…100, averaged inside each issue, then run through the same weights as the atlas. It is a placement, not a party match.

The second axis

Authoritarian (+) versus libertarian (−) is not “conservative versus liberal.” Gun rights score as liberty even when they sit on the American right. Speech restrictions, drug-war and sentencing hardness, immigration coercion, and curriculum mandates score as authority even when a left coalition backs them. That is why RFK Jr., Tulsi Gabbard, and Thomas Massie look like errors on a party map and like information on a compass.

Confidence, candidates, and 2026

Confidence is not a poll. It is a data-coverage number: more filled issues, plus a legislative score, plus incumbency, raise it. Candidates and cabinet officers are capped lower because they have not been forced to vote. A 2026 tag means the person is on a ballot this cycle, not that the score is a forecast of the race.

This atlas is a curated national set — executive, Senate, notable House members, governors, a few formers who still structure the argument — not all 535 voting members plus every down-ballot. Full-chamber scoring belongs to Voteview and GovTrack. We cite them instead of pretending to replace them.

Money on the profile

Each file shows net worth before the first office and now, plus this-cycle fundraising. Those numbers are modeled midpoints from House and Senate disclosures, Forbes estimates, FEC summaries, and published reporting — collapsed from ranges, not an audit of any household. A jump often is a spouse’s brokerage, a book, a company sale, or an inheritance, not a salary. Treat the bar chart as an argument about incentives, then read the note under it.

Stock holdings are STOCK Act / OGE ranges, collapsed onto tickers when the public file names them. A spouse’s brokerage counts. “Index funds” is itself a finding — most members do not day-trade, and the ones who do (Pelosi, Tuberville) are the story. Private-company fortunes (Hyatt, coal, dealerships, Carlyle) are listed as such, not dressed up as NVDA. This is not a live feed from anyone’s Robinhood.

Overview on a file is the one-paragraph read: who this person is when you strip the stump speech. It used to wear a cute name. It does not need one. College, spouse, children, and résumé sit as public-record bullets under the portrait — Wikipedia facts, not a campaign bio, and not the score. Salary is statutory pay for the seat they hold now (Congress frozen at $174,000 since 2009; cabinet on Executive Schedule I; governors at the state rate). It is not total compensation, not a pension, and not what the brokerage prints.

The live wire on each file polls the news as you sit on the page — about every twelve seconds — and a new article about that person lands at the top with a short summary and a link to the full piece. Wealth, holdings, and fundraising stay labeled snapshots; they do not tick with the market.

X sentiment is a seven-day read of posts about the person, scored as favorable / unfavorable / mixed and shown like a poll. It is not a scientific sample, not a registered-voter survey, and not the spectrum score. Open the posts. The bars are the room, not the census.

Odds are live implied probabilities from open Polymarket and Kalshi contracts on that person — not an endorsement, not a forecast we made. The number is what traders will pay for Yes. It refreshes while the file is open. Thin books move; read the volume.

City and county

The homepage line is national on purpose. Sitting mayors from Open States’ current municipal YAML, the 100 largest U.S. cities (Ballotpedia, August 2026), and elected county executives sit on each state’s page, on a second line. They are not mixed into the national atlas. Office email, phone, and address come from Open States. Short bios and portraits come from Wikipedia when the page is actually about that official — not a wrestler or a voice actor with the same name. County names come from official directories and the 2026 Maryland county-executive table. There is no DW-NOMINATE for a mayor. The score is the same twelve-issue model, seeded from party and a public record, with confidence cut. Most city elections are legally nonpartisan; those are coded independent unless a partisan identity is on the record.

Every council, sheriff, and school board in America is not here. Texas county judges and Kentucky judge-executives are not dumped in bulk — that would fake granularity the same way dumping 7,000 state legislators would. Wikidata’s live SPARQL endpoint rate-limited us to one query a minute. Wikipedia’s REST API did the same on a bulk pass. What you see is the free public file we can name and check: current mayors, the large-city overlay, and county executives with a single elected office.

Constituent reviews

Anyone can rate a figure and leave a short comment. The published name is first name plus last initial — Jane D., not Jane Doe. Email is required and stays off the public file. A verification code sits in front of the submit button so a script cannot flood the file. One review per person per week from the same connection. Readers can mark a review up or down — one vote per connection. This is not a poll, not a scientific sample, and not part of the spectrum score.

What this is not

• Not a poll of voters, donors, or staff.
• Not an endorsement, and not a moral ranking. Left and right are coordinates.
• Not live. Scores are modeled from the public record as of August 2026. People move. Fetterman is the exhibit.
• Not two-party destiny. Collins, Manchin, Massie, and Gabbard exist to break the habit of reading a D or an R as a number.