Skip to content

Do the experts really think RCV-IRV is "bad"? — grading a common claim

One of the most repeated arguments in electoral reform is that the voting-science community — mathematicians, game theorists, statisticians — is critical of instant-runoff voting (RCV-IRV) and favors cardinal methods (Approval, Score, STAR), while the general public wrongly treats RCV-IRV as the ultimate fix. This page takes that argument seriously and grades it: what's solidly true, what's overstated, and what it leaves out. It is not a verdict that RCV-IRV is "bad" — it's about how strong the popular version of the claim actually is.

Level: 201 → 301 · for debaters Prerequisites: What makes a "good" winner? · What makes a voting method good? · advocacy organizations.

The claim, in its forceful form

A common and confident version of the argument runs:

The public thinks RCV-IRV is the ultimate election fix, but if you look at the academics and statisticians who actually measure these systems, the math proves RCV-IRV is mathematically flawed and inferior to other models — which is why the experts favor Approval, Score, and STAR instead.

It's usually backed by naming figures like Warren Smith, Steven Brams, and Jameson Quinn (bios at the bottom). There's a real pattern underneath this — and two moves in it that don't survive scrutiny.

What's solidly true

  • RCV-IRV has genuine, provable flaws. It fails monotonicity — ranking a candidate higher can, in constructed cases, cause them to lose — and it is prone to center squeeze: a broadly-acceptable middle candidate can be eliminated early for lacking first-place votes, even when a majority would prefer them head-to-head. These aren't rhetoric; they're theorems with worked, re-runnable elections in this repo: the monotonicity before/after pair — 34 voters, where raising the winner X on four ballots makes X lose (X wins the baseline; Z wins after the raise) — plus Brams's IRV pathologies sampler and the center-squeeze set. IRV also isn't summable — precincts can't report simple subtotals — which complicates auditing.
  • Simulation studies do tend to rank cardinal/hybrid methods above RCV-IRV. Under VSE / Bayesian Regret the ordering is consistently roughly STAR ≳ Approval > RCV-IRV > Plurality. That is a real, repeatable result and a legitimate point in cardinal methods' favor.
  • Many theorists genuinely do prefer cardinal methods. The preference the claim describes is not invented — a substantial part of the social-choice and simulation community rates Approval/Score/STAR highly and is critical of IRV.

So the core observation — "the people who model these systems tend to be harder on IRV than the public is" — is largely fair.

What's overstated or missing

1. "The math proves it's inferior" is too strong. The math proves specific failures (monotonicity, center squeeze). It does not prove overall inferiority, because there is no single scale of "better":

  • Arrow's and Gibbard–Satterthwaite's theorems guarantee that no method satisfies every reasonable fairness criterion at once. "Best" is therefore always relative to which criteria you weight — a choice, not a proof. (RCV-IRV actually satisfies some criteria the cardinal methods fail; see below.)
  • VSE and Bayesian Regret are evidence under a model, not proof. Every such result is conditional on the assumed voter model and on how strategically voters behave — this repo's own simulation-models page opens by insisting "every conclusion is conditional on the model." Change the electorate model or the strategy assumption and the middle of the ranking can reshuffle. Robust as a trend; not a QED.
  • Utility itself is contested. VSE scores a method by summed voter utility — but whether one person's "5" is comparable to another's "5" (interpersonal utility comparison) is a real, old objection to the whole cardinal-utilitarian framing. It's a genuine argument for ranked ballots that the "the math proves it" framing simply omits.

2. The named "measurers" are also advocates. This is the biggest omission. Presenting Smith, Brams, and Quinn as disinterested referees is misleading in the same way an appeal to FairVote would be:

  • Warren Smith founded the Center for Range Voting — he is Score voting's leading advocate.
  • Steven Brams is the best-known academic champion of Approval voting.
  • Jameson Quinn (1974–2025) served on the boards of both the Center for Election Science (Approval) and the Equal Vote Coalition (STAR).

Their work is serious and worth engaging — but they are arguing for the methods their numbers favor, and the pro-IRV side (chiefly FairVote) produces its own favorable analyses. Advocacy pages from every camp mix solid points with overreach — including pro-STAR ones. "Cite the experts" cuts in more than one direction.

3. "Public vs. experts" is a loaded frame. Cardinal advocates are experts; so are the political scientists who study RCV-IRV's adoption, legitimacy, and real-world effects and weigh things axiom-counting ignores (ballot familiarity, the simplicity of the majoritarian story, a large track record). It isn't science-vs-ignorance; it's one expert camp vs. another, plus a public that mostly knows only RCV-IRV because it has the adoption momentum, not because anyone proved it best.

RCV-IRV's honest virtues (steelmanning the other side)

A fair page has to state what RCV-IRV does well, because the cardinal methods this repo favors don't:

  • Later-no-harm. In RCV-IRV, ranking a second choice can never hurt your first choice. STAR, Approval, and Score fail this — a high score for your backup can help it beat your favorite in the runoff or tally. For many voters this is the intuitively "safe" property, and it's a real cost of the cardinal trade-off.
  • The majority criterion, cleanly: a candidate who is the first choice of an outright majority always wins. (STAR meets this in spirit but has subtler majority behavior.)
  • Resistance to some strategies. Because it reads rankings, RCV-IRV doesn't invite the score-inflation / "min-max" incentives that pull Score voting toward Approval under pressure.
  • A large real-world record, which cardinal methods largely lack at government scale — an empirical, not axiomatic, point that legitimately matters to practitioners.
  • Monotonicity failures appear rare in realistic elections — how rare is debated, but "it fails monotonicity" is a possibility result, not a claim that real IRV elections routinely misbehave.

None of this makes RCV-IRV better than STAR on the simulation metrics — it doesn't. It makes "experts have proven IRV inferior" the wrong summary of a genuine, criterion-dependent disagreement.

The honest takeaway

Stated carefully, the defensible version is:

On the metrics voting scientists most often use — axiomatic criteria like monotonicity and center-squeeze resistance, and simulated voter satisfaction (VSE) — cardinal and hybrid methods (Approval, Score, STAR) tend to score better than RCV-IRV, and much of the theory community prefers them. That's a real point in their favor. It is not a proof that RCV-IRV is "bad," because "better" depends on which criteria you value, the simulations are conditional on their models, and RCV-IRV satisfies things (later-no-harm, a clean majority guarantee) that the cardinal methods give up. The experts most cited for the claim are also advocates — as is the other side.

That's the version this repo will stand behind: STAR looks strong here, honestly — and we say why the claim is strong without borrowing certainty the math doesn't supply.

The people named (accurately)

Person Background What they contributed Advocacy
Warren D. Smith Ph.D. in applied math, Princeton Bayesian Regret simulations; argued Score/Range outperforms RCV-IRV Founder, Center for Range Voting (Score advocate)
Steven Brams Game theorist, Professor of Politics, NYU Co-author of Approval Voting (1983); many papers on RCV-IRV paradoxes incl. monotonicity Leading academic Approval advocate
Jameson Quinn (1974–2025) Ph.D. in Statistics, Harvard (2020) Voter Satisfaction Efficiency (VSE) — the modern simulation metric; modeling favorable to STAR/Approval; also designed the Hugo Awards' E Pluribus Hugo nominating method Board member of both the Center for Election Science (Approval) and the Equal Vote Coalition (STAR)

In memoriam — Jameson Quinn (1974–2025). Quinn died in March 2025; VSE, his contribution, remains central to how methods are evaluated — this page included. That he was an advocate takes nothing from the work. → Full memorial

Naming them is fair evidence; naming them as neutral authorities is the move to resist. Read their work, note their affiliations, and check the claims against countable elections — which is the whole method of this repo.