Atlas / Newsroom / 2026-08-01

New open dataset finds America’s two parties further apart from each other than Republicans are from interwar fascism — and explains why that figure is misleading

The Political Atlas, published free today, codes 319 political organisations from 508 BCE to the present on six independent axes, with a citable source behind every one of 1,914 scores. Its first published analysis tests both of the accusations that define the American midterm cycle with the same instrument, and neither survives.

OTTAWA, Canada — 2026-08-01

A free research instrument published today maps 319 political organisations — parties, movements, regimes and doctrinal schools across twenty-five centuries — on six independent measurements rather than the single left-to-right line that dominates political argument.

Its first full analysis, published in its entirety alongside the dataset, applies the same instrument to both of the accusations that define the 2026 American midterm cycle. Neither holds. The Republican Party sits 16.0 from Italian Fascism on a scale where the maximum possible separation is 49.0. The Democratic Party sits 16.5 from classical Marxism. The two parties sit 18.8 from each other.

That last figure is the one the atlas warns against quoting alone. “The two parties are further from each other than the Republican Party is from interwar fascism, and that sentence is true,” the analysis states. “Anyone who repeats it without the next paragraph is misusing this document.”

The distinction the numbers make

The separation between the American parties is almost entirely about content — nine points on economic organisation, eleven on social order, ten on the rule of membership. The separation from fascism is almost entirely about form — ten points on method, nine on the scale of decision, six on the organisation of power.

“Those are not the same kind of gap, and a single number cannot tell them apart,” said Kenneth McGrath, who built the atlas. “The two parties disagree profoundly about what they want. What separates a democratic party from a fascist one is not what it wants — it is how it holds power, and that is a different measurement entirely.”

Both accusations fail, in different directions

The analysis records that since 2016 the Republican Party has moved on exactly one of the six axes — the organisation of power, by two points — with economics, social order, scale, membership and method all unchanged. Over ninety-two years the Democratic Party’s economic position has moved one point, while its position on the rule of membership has moved seven.

“Both sides of American argument are correct that the other has changed,” McGrath said. “Both are wrong about which axis. The entire dispute is conducted in economic language about movements that happened somewhere else.”

The atlas also codes the movement inside the Republican Party as a separate entry, which sits 10.1 from Italian Fascism — measurably closer than the institutional party. That entry carries the lowest confidence rating in the dataset and is flagged as contested, because specialists divide on which features define the term.

An instrument that publishes its own failures

Unusually, the atlas runs tests against itself on every page load and publishes the results, including those that weaken it. One reports that two of its six axes correlate at 0.86. Another finds that dropping any single axis leaves a distance matrix still correlating with the full model above 0.95 — meaning the six-axis model is not defensible on aggregate statistics alone, and is justified instead by a named twenty entries that a smaller model would misplace.

A further check, added after it found ten genuine errors, verifies that the example cases named in the codebook actually sit where the codebook says they do. The Pirate Parties had been cited as the cultural midpoint while scoring −07.

“An instrument that only ever reports findings in its own favour should not be trusted,” McGrath said. “The honest answer to why should I believe this is that you should not, until you have read the evidence. So all of it is published.”

What is available

The full dataset is free to download, reuse and redistribute with attribution, as JSON, and includes every score, the rationale and citable artefact behind each of the 1,914 scored cells, the trajectories of twenty organisations across historical anchor points, 135 recorded lineage relationships, and a logged history of every score that has been revised.

The atlas states plainly what it cannot do. It is not peer reviewed, it is not the output of the Chapel Hill Expert Survey, the Manifesto Project or V-Dem, and it has a single coder — which means it cannot report the inter-coder reliability figure that professional datasets publish. The method page names that as its largest weakness, and the site includes an instrument that lets any reader code a cell blind and compare their answer with the atlas’s.

Links

• The full analysis, published unredacted
• The dataset
• Method, limits and self-tests
• The codebook
• The interactive atlas

Notes to editors

• The Political Atlas is free to use, quote and reproduce with attribution. There is no paywall and no registration.
• Distances are straight-line separations across six axes each running from −10 to +10, giving a maximum possible separation of 49.0.
• Every figure in this release is reproducible from the published dataset.
• Interviews, custom comparisons and data queries are available to journalists on request, without charge.