Skip to content

FairVote's claims, checked — a standing ledger

FairVote is the largest and oldest US voting-reform organization, and its materials are the ones a STAR or Ranked Robin advocate is most likely to meet in an actual debate. This page is the running index of every FairVote claim this repo has checked against a countable election or a published theorem — the ones that hold, the ones that overreach, and the ones that are false. It is meant to keep growing; each row links to the full check.

Level: 301 · for debaters

How to use this page (read this first)

This is a ledger, not an attack page, and the distinction is the whole point. A page that collected only errors would be worth exactly nothing in a debate — the first true FairVote claim anyone quoted would sink it. What makes this list usable is that it records verdicts, including "holds," and that some of those verdicts land against STAR and stay on the page anyway. Three of the checks below concede a real STAR weakness; one runs FairVote's own strategic example and confirms it works.

Three rules this ledger follows, from the repo's fairness discipline:

  1. Quote before you check. Every row quotes FairVote's own wording. A paraphrase you can argue against is not a claim they made.
  2. Check against something that can say no — a runnable election or a theorem, never a preference.
  3. Disclose both leans. FairVote advocates RCV-IRV; this repo teaches STAR. Neither lean is a defect; concealing either one is.

And the standing tone rule for this repo: sing STAR's and Ranked Robin's virtues rather than attacking RCV. RCV voters are the natural allies of a better count — Ranked Robin uses their exact ballot. Use these checks to correct a specific claim once, then get back to what the methods do well.

The ledger

Verdicts: ✅ holds — accurate as written · ⚠️ overreaches — a true core stated too strongly, or a standard applied to one side only · ❌ false — checkably wrong.

Source Claim Verdict The check
Comparison chart (undated) "RCV … satisfies both the Independence of Irrelevant Alternatives and Independence of Clones criteria" ❌ Impossible by Arrow, and refuted by Burlington 2009 and Alaska 2022 — the two elections FairVote itself names as its Condorcet failures. Run the counterfactuals
Comparison chart "Plurality, approval, score, and STAR voting … fail … the Condorcet Loser Criterion" ❌ STAR passes it — the automatic runoff makes it structural. FairVote's own 2018 white paper concedes this. The row
Comparison chart "proportional analogs to Condorcet, approval, score … have seen … little study or advocacy" ❌ The literature starts with Thiele (1895) and Phragmén, has a 2023 Springer textbook, and a reference implementation this repo runs as an engine. The row
Comparison chart "strategic voting is not a concern in jurisdictions and among voters that use RCV" ⚠️ Later-no-harm and burial-immunity are real and conceded. But compromising is favorite betrayal, and Burlington 2009 is a live case — the same paragraph concedes it, then generalizes past it. The row
Comparison chart "These methods are vulnerable to the election of a candidate who lacks majority support" (of approval/score/STAR) ⚠️ STAR's runoff is a head-to-head majority; what it can fail is majority favorite, which is a different criterion. The failure is real, the stated consequence is not. The row
Comparison chart "the 'beats-all' winner only lost twice" in ~500 US RCV elections ✅ Real, checkable, and the best argument on the page. Our own caveat is about field size, not the number. Measured
Comparison chart "RCV and Condorcet methods are more complex than a simple arithmetic sum" ✅ FairVote grades its own method Low here. Correct — with one omission, summability, where RCV and Condorcet methods are opposites.
Comparison chart STAR/score/Condorcet "have never been used in a public election for government office" ✅ True, and the guinea-pig risk it names is a real political cost. The row
"Explaining FairVote's position on STAR Voting" (Richie & Penrose, 2018) STAR fails Later-No-Harm ✅ True and conceded — and it is the whole disagreement, not a slip. STAR's camp argues LNH is what forces center squeeze. Full check
2018 white paper STAR fails majority-favorite and mutual-majority ✅ True and conceded. STAR's honest limits
2018 white paper A coordinated faction can bury a centrist out of STAR's runoff (French 2017, Washington 2010) ⚠️ The burial works — we ran both. The overreach is what goes unsaid: it needs every rival faction to bury the centrist and rate its enemies a 4, and RCV-IRV on those same strategic ballots elects Le Pen. Both examples, tabulated
"Why the Condorcet Criterion Is Less Important Than It Seems" (Slatky, 2010) "Condorcet winners are centrist by nature, regardless of the preferences of the electorate" ❌ Self-contradictory — the Condorcet winner is defined by those preferences. Shift the electorate left and the pole candidate becomes the Condorcet winner. Run it (BV2169)
2010 Condorcet article "the impossibility of victory under Condorcet methods" for voters who lean to one side ❌ Non-moderates win whenever a majority prefers them pairwise. It is the compromise candidate who faces impossibility — under the first-choice methods. Run it (BV2170)
2010 Condorcet article Electing the pairwise winner is "the fallacy of the middle ground" ❌ Category error — the middle-ground fallacy is about the truth of propositions, not the aggregation of expressed preferences. The check
2010 Condorcet article No single ideal of a "best" winner exists, so "Condorcet isn't everything" ✅ The article's one genuinely good point, and this repo makes it too. What makes a good winner?

The pattern across all three documents

Not "FairVote lies" — something more specific and more useful to name in a debate:

  • The standard slides between rows. Spoiler resistance grades RCV on fixed ballots and cardinal methods on hypothetical voter behavior. A strategy is excused for RCV as too risky and coordination-heavy, and the same reasoning is not extended to STAR.
  • Criteria get conflated. Majority favorite is stated as "lacks majority support." IIA is described in prose and clone-independence is cited as the proof. Burial (a strategy) is used to grade the spoiler row, which has its own row two criteria earlier.
  • The best claims are empirical, and they hold. The 2-in-~500 Condorcet figure, later-no-harm, STV's record, the representation research — these are strong and should be conceded first in any exchange. The errors cluster in the mathematical claims, which is exactly where they are checkable.

Adding a row

New check? Follow the recipe rather than improvising it — it is what makes this list quotable:

  1. Quote it verbatim, with the document and a retrieval date.
  2. Build the smallest election that settles it (keep examples small), or cite the theorem that does.
  3. Run it and paste a generated report block — never a hand-typed table (why).
  4. Write the verdict against yourself first. If the claim holds, say so and mark it ✅. A ledger with no ✅ rows is not evidence, it is a grievance.
  5. Add the row here and put the full check on a page of its own.