Bad" method, and how it compares to STAR¶

A well-regarded rated method designed by Jameson Quinn (Center for Election Science). Voters rate every candidate Good, OK, or Bad; the winner is found in three steps — 3 semifinalists, then 2 finalists, then 1 winner. It dodges the same strategic traps STAR does (no center squeeze, a non-slippery chicken dilemma, a built-in dark-horse guard), which is why it keeps coming up as a serious reform option. This page explains it fairly and lays it next to STAR — they're close cousins, with real trade-offs either way.

The ballot and the three steps¶

Ballot: rate each candidate Good, OK, or Bad (a 3-level rated ballot — coarser than STAR's 0–5, finer than Approval's yes/no).

3 — Semifinalists. Take the three candidates with the most Good ratings.
2 — Finalists. Of those three, keep the two with the fewest Bad ratings.
1 — Winner. Of those two, the winner is the one rated higher on more ballots — the pairwise winner (Good > OK > Bad).

Two guard rules on the third semifinalist (they're what make 3-2-1 more than a toy):

Clone guard: the third semifinalist can't be a near-duplicate of the first two (not the same party as both, or — nonpartisan — don't count its "Good"s on ballots that also rated the first semifinalist Good). Stops a party from cloning its way into the finals.
Dark-horse guard: the third semifinalist must have at least ½ as many Good ratings as the top semifinalist. If none qualifies, skip step 2. This is what keeps a dark horse — a nobody with almost no real support — out of the finals.

Worked example — it resolves the chicken dilemma (usually)¶

Take the chicken/Burr scenario: allies A and B must beat C, whom the majority opposes (35 prefer A>B, 25 prefer B>A, 40 back C). If the A/B voters use OK to separate their favorite from their ally (A-voters: A Good, B OK; B-voters: B Good, A OK; C-voters: C Good, both allies Bad):

             Good   Bad
   A          35     40
   B          25     40
   C          40     60
Step 1  → semifinalists A, B, C (only three candidates)
Step 2  → finalists A, B        (fewest Bad; C's 60 Bad knocks it out)
Step 3  → A beats B, 35 to 25   → A wins

3-2-1 elects A — the honest pairwise winner — and the majority-opposed C is knocked out in the finalist step. Same answer STAR gives, and no slippery slope.

The honest catch — granularity. If those same A/B voters rate both allies Good (both are genuinely liked, after all), then no voter distinguishes A from B, step 3 is a 0–0 tie, and 3-2-1 falls back to a coin flip — a milder echo of Approval's chicken tie. 3-2-1 only cleanly resolves the finals when voters spend the OK level to rank their top two. STAR's 0–5 ballot sidesteps this: scoring A 5 and B 4 is natural and low-stakes, so the runoff almost always has a real preference to read. This is the preference-vs-support / scale-granularity point in miniature: more levels means the final step can actually decide.

What 3-2-1 is good at¶

No center squeeze — the finalist step is by fewest-Bad, not fewest-first-choices, so a broadly-liked centrist isn't eliminated for lacking first-place love the way IRV does.
Non-slippery chicken — like STAR, a few defectors can't start an avalanche (a final head-to-head means small defections don't snowball).
Explicit dark-horse and clone guards — where STAR resists dark horse implicitly (via support-reading), 3-2-1 bolts on rules that do it explicitly.
Genuinely explainable — "your worst candidate gets knocked out before the finals" is an intuitive, even chantable pitch; several people find it the easiest good method to explain across the aisle.
Summable — precincts can report Good tallies, Bad tallies, and the pairwise matrix; the winner is computable from those aggregates (no need to centralize raw ballots). Same virtue STAR has.

3-2-1 vs. STAR — close cousins, real trade-offs¶

	STAR	3-2-1
Ballot	0–5 (6 levels)	Good / OK / Bad (3 levels)
Tabulation	add scores → top 2 → pairwise runoff (2 steps)	most-Good → fewest-Bad → pairwise (3 steps + 2 guard rules)
Center squeeze	avoids	avoids
Chicken dilemma	non-slippery	non-slippery (but see the tie caveat)
Dark horse	resisted implicitly (reads support)	resisted explicitly (½-Good guard)
Summable	yes (score totals + matrix)	yes (Good + Bad tallies + matrix)
Granularity	finer — breaks near-ties in the runoff	coarser — top two can tie unless voters use "OK"
Spec complexity	simpler to state — no special-case rules	needs the clone / dark-horse rules to behave
Backed by	Equal Vote Coalition	Center for Election Science / Quinn

Neither is strictly better. STAR is finer-grained and structurally simpler (two clean steps, no special cases), and its extra levels break the exact tie that can trip 3-2-1. 3-2-1 answers voters who find 0–5 daunting with just three words, and it hard-codes the dark-horse and clone protections that STAR gets for free from its scores. Both are honest, summable, center-squeeze-free rated methods — worlds better than plurality or IRV, and disagreeing mainly at the margins.

Honest limits¶

Not strategy-proof — Gibbard forbids that for any method. Quinn notes 3-2-1 becomes tactical with four or more serious candidates.
The guard rules are real complexity — the "third semifinalist" clone and dark-horse conditions are fiddly to write into law and to explain in full, even if the top-line pitch is simple.
Three levels can under-resolve — the finalist tie above; a Good/OK/Bad ballot simply carries less information than 0–5.
Not in public use — 3-2-1 is a proposed method (no governmental adoption we know of) and is not offered by BetterVoting, so this library can't back a 3-2-1 case with a live election the way it does STAR / Approval / Ranked Robin / IRV. The example above is worked by hand; the LH engine does not tabulate 3-2-1.

Sources¶

3-2-1 voting — electowiki — the clearest mechanics reference for this niche/branded method (advocacy-adjacent: electowiki leans reformer, and 3-2-1 is a Center for Election Science proposal — cite it for how it works, not for verdicts).
Jameson Quinn, The Six Voting Molochs — where 3-2-1 is proposed as the non-slippery fix to the chicken dilemma; Quinn is CES-aligned (Approval/3-2-1), which is exactly why his praise of methods he didn't design carries weight.
Contrast within this library: STAR · Approval · Ranked Robin · glossary 3-2-1 voting.