Skip to content

Voting 401 — failure modes & the safety check

Audience: anyone about to choose or adopt a method — officials, reformers, committee chairs, and the seriously curious. You've learned how the methods work (101), read their results and properties (201/301); this is the responsible last step before you'd stake an election on one: walk its failure modes with clear eyes. There is no perfect method, so the skill isn't finding a flawless one — it's knowing exactly which whoops you're signing up for, and how often they actually bite.

Prereq: 101 + 201 + 301. · Up: Curriculum hub.

The one idea: every voting method fails some reasonable criterion — that's a theorem, not an opinion. 401 is the safety check: for the method you like, find out how it fails, whether the failure is real-world or lab-grade, and whether you can live with it.

401.1 — Why there's no perfect method (the frame)

401.2 — The tool: reading a claimed failure honestly

  • Objective: be able to weigh any "method X fails!" claim — including the ones aimed at the method you favor.
  • Key terms: the four-part test — structural / sincere / realistic / really-happened.
  • Material: Reading these fairly — the test for an honest "whoops" and criteria at a glance.
  • The point: a failure that is structural, triggered by sincere votes, on a realistic electorate — and that has actually happened — is a serious mark against a method. A contrived, strategic, knife-edge construction never seen in the wild is the weakest, muddiest angle (and applies to every method via Gibbard). Grade every whoops on that scale.

401.3 — The safety datasheets (each method's honest limits)

  • Objective: read the conceded failure list for each method straight from its own honest-limits page — the fair 401, method by method.
  • Material:
  • STAR → STAR's honest limits (the why) and every criterion STAR fails (the index, with a runnable example + BV id for each): not Condorcet-compliant; trades Later-No-Harm; rare majority / participation / mono-raise-delete / IIA / consistency failures.
  • Ranked Robin / Condorcet → RR's honest limits (cycles; blind to intensity — the sincere dark horse).
  • RCV-IRV → non-monotonicity, center squeeze, exhausted ballots.
  • Approval / Score → the chicken/Burr dilemma and the "where do you draw the line?" threshold question (are equal-score votes discounted?).
  • Plurality / Choose-One → the spoiler effect and wasted votes — the failures the whole reform movement exists to fix.
  • The point: a method that lists its own limits out loud is more trustworthy, not less. Candor is the credibility (it's 301's rule, carried to the capstone).

401.4 — The failures that actually happened

  • Objective: separate real-world failures from lab constructions — the top of the four-part test.
  • Material: Burlington 2009 (center squeeze + non-monotonicity on certified ballots → IRV repealed), Alaska 2022 (Condorcet failure, spoiler, upward monotonicity, exhaustion — one real race), and San Francisco D7 2020 (downward monotonicity).
  • The point: these are IRV's, and they're structural and sincere — which is why they carry weight that a contrived STAR gotcha doesn't. Keep the rarity in view too (2 Condorcet failures in 182 US RCV elections): real, predictable, but not constant.
  • Objective: see each named criterion fail on a real, re-runnable election — the concrete counterpart to 401.1's theorems.
  • Material: monotonicity (raise the winner → they lose), reversal symmetry (best = worst), the Dark Horse (Borda elects a nobody), the chicken dilemma, and favorite betrayal.
  • The mean twin: the Mudroom collects these as deliberately unfair, one-sided per-method hit-reels — fun, clearly labeled, and not for citing. Use it to appreciate that no method is spared; use the fair pages above for the actual analysis.
  • The point: the impossibility theorems say failures must exist; these make them tangible — and, seen across every method, prove the failures are universal, not a mark of one "bad" method.

401.6 — Reading the advocacy (claim-checking)

  • Objective: apply the safety check to the literature, not just the methods — every camp's white paper leans, and each is worth checking with the ballots in hand.
  • Material: the worked claim-checks — FairVote's official position on STAR (pro-RCV), the FairVote Condorcet article, and the rangevoting.org "IRV logic" critique (pro-Range). Each concedes the valid core, runs the testable claims, flags the overreach, discloses the lean — and lets the fairness cut against STAR too.
  • The point: the same discipline that grades a method's failure grades an argument's: concede the kernel, test it, name the overclaim, disclose who's talking.

401.7 — The synthesis: choosing responsibly

  • Objective: turn the failure inventory into an actual decision.
  • Material: what makes a good winner?, what makes a voting method good?, and the VSE foundations in 301.
  • The point: weigh the failure modes by severity × frequency, not by count. A method with a rare, lab-only seam beats one with a common, sincere, real-world one — even if a checklist shows each "failing" something. STAR and Ranked Robin trade a few contrived, seldom-seen failures for the elimination-order disasters IRV suffers sincerely; all three tower over Choose-One. That comparison — made honestly, with the whoops on the table — is the whole point of the library, and the reason to run the safety check before you adopt, not after.

Up: Curriculum hub · Prev: Voting 301. The unfair, for-sport companion to this page is the Mudroom — same failure modes, no bedside manner.