The spoiler effect¶
A spoiler is a candidate who can't win but still changes who does — by splitting another candidate's support. It's the single most-felt flaw of choose-one voting, the reason "don't waste your vote on a third party" is common advice, and the problem every reform on this site is partly trying to fix.
→ Glossary: spoiler effect · the root cause: vote splitting & the equally-weighted vote · runnable demos: the split-voting set · how often it happens for real: the 2022 count, read honestly · where it sits among the other ways a vote loses its power: the problem, in four causes
What it is¶
You have three candidates: two similar ones (say two progressives) and one different (a conservative). Most voters prefer either progressive to the conservative — but choose-one forces each voter to pick exactly one. The two progressives divide their shared supporters, so neither reaches the top, and the conservative wins with a minority. The weaker progressive was the spoiler: they couldn't win, but their presence flipped the result.
Nothing about the electorate's actual preferences changed — only the ballot's refusal to let a voter support more than one candidate. That's why the spoiler effect is a property of the counting rule, not of voters being fickle.
The root cause: one mark¶
The spoiler comes from vote splitting — similar candidates sharing one pool of supporters who are each allowed only one mark. Fix the ballot so a voter can back several candidates at once, or express degree of support, and the forced split goes away. This is the same root the equally-weighted-vote argument starts from.
How each method handles it¶
| Method | Spoiler behavior |
|---|---|
| Choose-One / Plurality | Full spoiler effect — the classic case above. |
| RCV-IRV | Reduces the classic spoiler: a trailing similar candidate is eliminated and their ballots transfer, instead of splitting the vote outright. But elimination on first-choices creates its own failure — center squeeze — and can still spoil in some configurations. |
| Ranked Robin / RCV-RR | Reads the whole ballot and elects the head-to-head (Condorcet) winner — no elimination rounds, so no vote splitting or center squeeze. Only a rare Condorcet cycle can produce a spoiler. |
| Approval | Approve both similar candidates — they no longer split, since a supporter can back all of them. |
| Score / Range, STAR | Score each candidate independently, so running an ally doesn't bleed your support. STAR adds a majoritarian runoff on top. |
How often is the setup there?¶
Constantly. Of all 5,662 US national primaries in 2022 — congressional, statewide, state legislative — about 11.9% had multiple candidates and a winner without a majority, meaning the losing candidates' voters together outnumbered the winner's (Center for Election Science, America (Mis)Represented). That is the exact configuration a spoiler needs.
It is not, however, a count of spoiled elections — a sub-majority winner can still be the candidate a majority genuinely prefers, and under a one-mark ballot there is no way to tell which case you're in. Both readings, and two runnable counter-examples, are in How often does vote splitting actually happen?.
Spoiler ≠ center squeeze¶
These get conflated, but they're different failures:
- Vote splitting / spoiler — similar candidates share one pool of supporters (the progressives above). The fix is letting voters support more than one.
- Center squeeze — a distinct moderate, liked by both sides, is eliminated by RCV-IRV for too few first choices even though they'd beat each rival head-to-head. The voters here are different, not shared. This one is IRV-specific.
Does STAR fully escape it?¶
Almost. STAR removes the forced split, but a narrow, self-inflicted residual survives in the top-two runoff — a faction can still split itself if it refuses to use the score ballot honestly. That edge case (and why it's minor) is STAR's residual vote-splitting.
How social choice theory frames it¶
A method is spoilerproof when it satisfies Independence of Irrelevant Alternatives (IIA): whether the winner is A or B should not depend on some losing third candidate C. Sidney Morgenbesser's joke captures the principle — a diner picks apple pie over blueberry and cherry; told the cherry is sold out, he says "in that case I'll have the blueberry." Condorcet was the first to study the spoiler effect, in the 1780s.
Susceptibility depends heavily on the method (Wikipedia's summary):
| Electoral system | Spoiler effect |
|---|---|
| Plurality | High |
| Runoffs / RCV-IRV | Medium |
| Condorcet methods (Ranked Robin / RCV-RR, Schulze, Ranked Pairs) | Low (only in a cyclic tie) |
| Score / median (Approval, Score, STAR) | None (on an absolute scale) |
Two theorems sit under this. Arrow's impossibility theorem shows no ranked method can fully escape spoilers; rated methods aren't subject to Arrow, so they can be spoilerproof — with one caveat: only if voters score on an absolute scale. If voters rescale their scores around who's running (a relative scale), a spoiler can reappear — which is exactly STAR's narrow residual.
Methods that fail IIA can be gamed by strategic nomination. The Borda count has a severe entry incentive — a faction can "clone their way to victory" by running many candidates, which made Borda himself concede "my system is meant only for honest men." Plurality has the opposite, an exit incentive: similar candidates get pressured to drop out or merge so they don't split the vote. (Theory suggests 90–99% of real elections have a Condorcet winner, so cyclic-tie spoilers are rare; the first Condorcet cycle found in a ranked US election was in 2021.)
The rhetoric: "a vote for X is a vote for Y"¶
The spoiler effect has its own political vocabulary, all of it aimed at pressuring voters not to "split the vote": "a vote for X is a vote for Y." The line recurs across decades — "a vote for Perot is a vote for Clinton" (1992); Al Gore's campaign on "a vote for Nader is a vote for Bush" (2000, and again in 2004); "a vote for Cruz is a vote for Trump" in the 2016 GOP primary; and a 2024 Democratic ad branding Jill Stein a spoiler for Trump. This is lesser-evil / strategic-voting pressure — the human cost of a spoiler-prone ballot — and defusing it is much of the point of reform: on an Approval, Score, or STAR ballot you can honestly support your favorite and a viable compromise, so "a vote for X is a vote for Y" stops being true.
A real IRV spoiler — Burlington 2009¶
The 2009 Burlington, Vermont mayoral election (run by RCV-IRV) is the textbook proof that instant runoff does not remove spoilers. Andy Montroll was the Condorcet winner — he beat every other candidate head-to-head — but he was eliminated in an early round for too few first choices. Republican Kurt Wright acted as the spoiler, and the seat went to Bob Kiss, even though most voters preferred Montroll to Kiss. That's a center-squeeze spoiler, and it's why IRV rates "Medium," not "None," in the table above.
The most prominent recent example is Alaska's 2022 U.S. House special election, where RCV-IRV eliminated the Condorcet winner (Begich), a losing candidate (Palin) spoiled the result, and the outcome was even non-monotonic — three IRV pathologies in one real federal race.
See it happen¶
- The split-voting set — the spoiler progression from plurality through each reform, on the same ballots.
- Star Wars vote split — a compact worked example (Skywalker & Leia split the Rebel vote).
- Plurality vs. majority — where the minority winner comes from.
- Debate framing: what's so good about STAR — Segment 1
Sources: Spoiler effect — Wikipedia, Spoiler effect rhetoric — Wikipedia, Independence of irrelevant alternatives — Wikipedia, 2009 Burlington mayoral election — Wikipedia