Skip to content

The Ranked Ballot

One ballot style, many counts. A ranked ballot asks one question — "put the candidates in order" — and that single design choice fixes what any method downstream can ever know about you: your order, never your strength. This page is the anatomy of the ballot itself; its twin is the score ballot, and the side-by-side comparison is alternate ballot styles.

→ Companions: scores vs. ranks (the core distinction) · strict vs. weak ranks · the ranked-ballot method zoo (fifteen ways to count this one ballot) · terminology canon: TIPS_terminology

What it looks like

A ranked ballot: five candidates (Andre, Blake, Carmen, David, Ella) each on a row, with bubble columns 1st through 5th. The voter has marked Andre 1st, Carmen 2nd, David 3rd, Blake 4th, Ella 5th.

One row per candidate, one column per rank. This voter says: Andre first, then Carmen, then David, then Blake, and Ella last.

It's the same voter, the same opinion, as on the score ballot — line the expressions up and you can see what each style keeps:

Candidate Ranking (this page) Yes/No Score 0–5
Andre 1st ● 5
Blake 4th ○ 1
Carmen 2nd ● 4
David 3rd ● 4
Ella 5th ○ 0

Notice the middle of the field: the voter feels Carmen = David (both 4s), but this ballot has to call one of them 2nd and the other 3rd. The full side-by-side walk-through is alternate ballot styles.

The grid rules — and how they bite

A ranking is a grid constraint: on the common strict ballot, one mark per row and one per column. That coupling between rows is what makes the ranked ballot fussier to fill out than it looks:

  • Duplicate ranks (an "overvote") — mark two candidates 2nd and, in most US RCV-IRV rules, that portion of your ballot (or all of it) is void. You weren't allowed to feel a tie.
  • Skipped ranks — jump from 2nd to 4th and jurisdictions differ: some ignore the gap, some stop reading your ballot right there.
  • Rank caps — many real ballots only offer 3–5 rank columns in a large field. Whatever you couldn't rank, the count treats as silence — a ballot-format path into exhausted ballots.

These aren't hypotheticals: reported spoilage runs roughly 4–9% for ranked ballots vs. 0–2% for rated ballots (see scores vs. ranks), and the errors fall hardest on voters with the least practice navigating grids.

What it captures — and what it throws away

A ranking records relative order only. It cannot say how much you prefer 1st over 2nd, and it spaces every rung equally even when the voter feels a cliff: this voter's real feeling might be "Andre, Carmen, David all great; Blake and Ella, no" — but 1st/2nd/3rd/4th/5th flattens that cliff into a smooth staircase. Any method that later needs strengths has to invent them (that's Borda, and why rank→score conversion fabricates information — the fidelity ladder).

Strict vs. weak is the one ballot-design fork inside the ranked family: a weak ranking allows equal ranks (A = B), a strict one forbids them. Ranked Robin happily reads weak rankings; typical US RCV-IRV ballots are strict, which is exactly the "forced to invent a preference" problem. Full story: strict vs. weak ranks.

One ranked ballot, many tabulations

This is the repo's core terminology point: RCV names the ballot, not the count. The identical grid above can be tabulated by:

Count it with Family Notes
RCV-IRV (Hare) elimination rounds what "RCV" usually means in the US
Ranked Robin Condorcet / round-robin reads every pairwise matchup; allows equal ranks
STV proportional multi-winner transfers (STV vs proportional STAR)
Borda, Bucklin, and a dozen more positional & hybrids the method zoo runs 15+ on one screen

Different tabulations of the same ballots routinely elect different winners — which is why "the ballots decided" is never the whole story. See what is a voting method? — a ballot and a count.