FairVote's official position on STAR, claim-checked¶

In July 2018 FairVote — the leading Ranked-Choice-Voting advocacy group — published a white paper, "Explaining FairVote's position on STAR Voting" (Rob Richie & Drew Spencer Penrose), laying out why it backs RCV over STAR. It's the canonical opposition document, and most of its claims are checkable. This page quotes each, concedes the ones that are true (STAR does fail some criteria — and here's why we think the trade is worth it), runs the strategic examples on a live engine, and flags the ones that were already false when the ink was wet. Same discipline as the FairVote Condorcet claim-check — and, as there, the fairness cuts toward our own side too.

▶ Live on BetterVoting: French 2017 honest ↗ · burial ↗ (BV2229/2230) — Washington 2010 honest ↗ · burial ↗ (BV2231/2232).

The claims that are TRUE — conceded¶

FairVote is a serious critic, and several of its points are simply correct. Pretending otherwise would fail this repo's own fairness rule.

STAR fails Later-No-Harm. True — scoring a second candidate can help defeat your favorite. But this is the whole disagreement. FairVote lists LNH-failure as a flaw; STAR's camp — including STAR's co-inventor Clay Shentrup, "Later-no-harm is a bug, not a feature" — argues LNH is the very property that forces center squeeze: guaranteeing your backup can never help your favorite is mathematically the same as guaranteeing a broadly-liked compromise can't be rescued. RCV keeps LNH and pays for it in Burlington and Alaska. Worked in full on the favorite-betrayal page.
STAR fails the majority-favorite and mutual-majority criteria (in constructed cases); RCV satisfies both. True and conceded — see STAR's honest limits and residual vote-splitting. These are rare, constructed failures, not everyday behavior.
There is no single "honest" 0–5 score the way there's an honest ranking. A fair philosophical point — scoring asks how much, which is genuinely a judgment call. The repo agrees the scale is a modeling choice (are equal-score votes discounted?); the counter is that real STAR elections show ~90% of voters using the full range comfortably (STAR history).

And FairVote makes one concession in STAR's favor worth keeping: STAR cannot elect the Condorcet loser (the runoff prevents it), unlike plurality, score, and approval. They're right, and it's a real STAR strength.

The strategic examples — RUN, not just argued¶

The white paper's most concrete claim is that STAR's strategy incentives let a coordinated faction bury a strong centrist out of the runoff. It gives two examples. We modelled both — honest ballots and the exact burial FairVote describes.

The 2017 French election¶

FairVote: with four ~20% candidates and Macron the broadly-preferred centrist, rival factions could score Macron 0 (and inflate the other wings) to keep him out of the runoff.

	STAR	RCV-IRV
Honest (`french_2017_honest`)	Macron (351; runoff 51–49)	Macron
Coordinated burial (`french_2017_strategic`)	Mélenchon (Macron → 130, squeezed)	Le Pen

Conceded: the burial works under STAR — Macron, the Condorcet winner, is squeezed out. But read the honest full picture: it required every rival faction to bury Macron and rate its ideological enemies a 4 — a risky, wildly-coordinated conspiracy. Honest STAR elects Macron. And RCV-IRV on the same strategic ballots elects Le Pen — a worse miss. The asymmetry the white paper never states: STAR squeezes the center only under coordinated strategy; IRV squeezes it under sincere voting (Burlington, Alaska).

The 2010 Washington State Senate race¶

FairVote: anti-Berkey groups boosted the long-shot conservative Rieger to squeeze the moderate Berkey; under STAR they'd score Harper 5, Rieger 4, Berkey 0.

	STAR	RCV-IRV
Honest (`wa_2010_honest`)	Berkey (385; runoff 60–40)	Berkey
Coordinated burial (`wa_2010_strategic`)	Harper (Berkey → 225, squeezed)	Berkey — resists!

This one cuts fairly against STAR, and we keep it that way: the burial works under STAR (Berkey squeezed, Harper wins), and RCV-IRV on the same ballots still elects Berkey — the manipulation STAR fell for, IRV happened to resist. That's a real point for IRV in this case. The honest balance: (1) honest STAR elects Berkey; (2) the burial needs the Harper faction to rate a conservative a 4, which could backfire and elect Rieger; (3) IRV's resistance here is incidental to elimination order, and doesn't rescue it from squeezing centrists when voters are honest. One method needs a conspiracy to squeeze the center; the other does it sincerely.

The claims that were already OVERCLAIMED¶

"RCV elects the Condorcet winner in practice." The paper leans on "over 100 Bay Area RCV elections, every one elected the Condorcet winner." But Burlington 2009 — a real RCV Condorcet failure — predated this 2018 paper by nine years, and the claim is carefully scoped to the Bay Area. That's a criterion-built-to-fit tell; Alaska 2022 later refuted it outright. RCV does not satisfy the Condorcet criterion, in theory or in the field — a fact available when the paper was written.
The "nullifying ballots" framing. Equal-Support voters (who score both finalists the same) are described as having their votes "set aside." That's precisely the framing our are equal-score votes discounted? page rebuts: their scores are counted (they helped choose the finalists); they simply decline to tip a pairing they rated equally — exactly like leaving a down-ballot race blank.
"STAR has no track record." True in 2018; dated now. STAR has run binding public elections since the 2020 Independent Party of Oregon primary (history).

Net assessment¶

FairVote's white paper is a good-faith, serious critique — the best-argued opposition to STAR there is — and its criterion claims are largely accurate: STAR does fail Later-No-Harm, the majority-favorite criterion, and mutual majority, and it can be strategically manipulated to squeeze a centrist. We concede all of that. The disagreement is threefold: (1) whether LNH is a virtue — FairVote assumes yes; STAR's whole design says it's the bug that causes center squeeze; (2) whether STAR's strategic vulnerability is comparable to IRV's sincere one — it isn't, and the French/WA runs show honest STAR electing the centrist both times; and (3) the paper's one empirical claim, that RCV reliably elects the Condorcet winner, was already contradicted by Burlington when written and is now refuted by Alaska. Read every white paper — FairVote's and STAR's — with the ballots in hand.

The demo elections¶

All four are LH-verified and live on BetterVoting.

Case	Winner	Live results	Page
French 2017 — honest	Macron	BV2229 ↗	page
French 2017 — burial	Mélenchon	BV2230 ↗	page
Washington 2010 — honest	Berkey	BV2231 ↗	page
Washington 2010 — burial	Harper	BV2232 ↗	page

Both scenarios are simplified teaching models of the real fields (French: the ~20%-each four-way; Washington: Berkey / Harper / Rieger), built to reproduce the exact dynamic FairVote's white paper describes. Source of the critique: Richie & Penrose, FairVote, July 2018 — an advocacy document, checked here the same way this repo checks its own side's advocacy.