STAR's Hybrid Nature — Expressive Scoring + a Majority Runoff

One line: STAR is deliberately a hybrid. It takes the expressiveness of a score ballot (say how much you like each candidate, not just an order) and pairs it with the legitimacy of a majority runoff (the winner is the one more voters actually prefer between the top two). Score to find the finalists; majority to pick the winner.

In the formal language of the field, that means STAR combines a cardinal ballot (scores — how much) with an ordinal decision (the head-to-head runoff — which one). As Clelland (2023) puts it, STAR is "a combination of cardinal and ordinal voting systems," first introduced in 2014 — (arXiv:2303.00108; she uses it to explain why STAR gave a robust result on the Alaska 2022 ballots where pure-cardinal Approval did not).

→ The payoff this design produces: Runoff Reversal and Three notions of "winner". Glossary: STAR.

---

The two steps, in plain language

STAR = Score Then Automatic Runoff.

1. Scoring Round. Score each candidate from 0 to 5 stars, like rating products online. Add up everyone's stars. The two highest-scoring candidates become the Finalists.
2. Automatic Runoff. Now compare only those two finalists. Each ballot counts as one full vote for whichever finalist it scored higher. The finalist preferred by more voters wins.

Two steps, both things you can do on paper, and nothing is thrown away: every ballot is read in full, in both rounds.

Why combine the two ideas?

Each "pure" approach has a known weakness; STAR's hybrid covers both.

• Pure scoring (Approval / Score Voting) is wonderfully expressive, but with only a sum it can hand the win to a candidate adored by a passionate minority over one liked by a majority — intensity can outvote numbers.
• Pure majority / ranked methods respect numbers, but a bare ranking throws away how much you prefer one candidate to the next, and (in the case of RCV-IRV) the elimination machinery brings its own pathologies.

STAR's move is to use each idea for the job it's good at: scores measure support and surface the two genuine front-runners; the runoff then applies a majority check so the winner can't be a minority-intense pick. You get strength-of-support and a majority finish on a single, simple ballot.

In the vocabulary of what makes a "good" winner, that's a deliberate blend of two ideals: the scoring round targets the utilitarian winner (highest total support — the broadly-liked candidate), and the automatic runoff confirms it against the majority winner (preferred head-to-head). Score to nominate; majority to decide. Neither pure ideal alone lands there — which is the whole point of the hybrid.

The hybrid in one tiny election

Three voters, three flavors:

  Count ×  Almond  Brownie  Cocoa
      1 ×    5       1        2
      2 ×    4       5        0     (collapsed: appears twice)

→ run runoff_overturns_leader/01a_c3_b3_more-stars-fewer-voters.yaml

• Scoring Round finds the finalists: Almond (13 stars) and Brownie (11).
• Automatic Runoff picks the winner: 2 of 3 voters prefer Brownie, so Brownie wins 2–1 — even though Almond had more total stars.

That's the hybrid working as designed: scores nominated the two finalists; the majority runoff chose between them. Neither half alone would land there — pure score would crown Almond, and a method with no scores couldn't tell that Brownie was the broadly-liked compromise.

Where this leads

• When the scoring leader and the runoff winner differ, that's Runoff Reversal — the most important consequence of the hybrid design.
• When Condorcet, Score, and Runoff point at three different candidates, see Three notions of "winner".
• For where STAR sits relative to ranked methods, see Scoring methods aren't RCV and the step-by-step STAR vs RCV-IRV count.

Sources: STAR Voting (Wikipedia), Equal Vote Coalition — STAR Voting.