Skip to content

Is RCV "Simple"? — Which Half?

One line: ranking 1-2-3 really is simple to mark — but "simple" then gets quietly borrowed for the whole method, when IRV's count is the opposite of simple. STAR keeps both halves simple — and its ballot carries strength of support on top of order.

Applies to: the count of every sequential-elimination variant (Hare and the rest run multi-round elimination); the ballot is simple, the tabulation isn't. See Which RCV-IRV?.

→ Builds on What Is a Voting Method? — Ballot vs Count, Plurality vs Majority (a method = ballot + count) · related: IRV Is Not Summable — Every Ballot Must Be Counted Centrally, IRV Non-Monotonicity — When More Support Makes You Lose · step-by-step trace of both counts: How the Count Works — STAR vs RCV-IRV, Step by Step · Level 201.

The claim, and what's true about it

RCV advocates say: "It's simple — look at the ballot, just rank 1-2-3." And they're right about the ballot. Marking a ranked ballot is intuitive; ranking is the easy part. Concede that cleanly — you lose nothing.

The slip is in the next step. "Simple" is really two different claims, and only one of them is true:

  • Simple to mark — the ballot. True of ranking. (Also true of STAR.)
  • Simple to count and trust — the tabulation. This is where "simple" gets borrowed for the whole method when it was only ever earned by the ballot.

A method is a ballot and a count (see the 101 page). "Look how simple the ballot is" tells you nothing about the count — and the count is where the winner actually comes from.

→ This page argues the point for IRV; the cross-method version is a table. Every method placed on both axes at once — where Choose-One, Approval, RCV-IRV, 3-2-1, STAR and Ranked Robin each land — is in What makes a voting method good? §2.

Why IRV's count is the hard half

To find an IRV winner you can't just add a column. You:

  1. tally everyone's top remaining choice;
  2. if no one has a majority, eliminate the lowest and transfer each of those ballots to its next choice;
  3. recheck, and repeat — round after round — until someone clears half.

To follow it you have to track every transfer across every round. Three consequences, all working against "simple":

  • Not summable. A precinct's IRV result doesn't compose with another's — you can't add subtotals, because who's eliminated depends on the whole electorate. So every ballot has to travel to one central count. (See IRV Is Not Summable — Every Ballot Must Be Counted Centrally.)
  • Software-dependent and hard to hand-audit. Rounds of transfers are not something poll workers reconcile on a precinct table; a recount means re-running the algorithm on all ballots at once. (What that costs operationally and for public trust — courier runs, single point of failure, the real incidents: Central tabulation — when every ballot must travel.)
  • Counterintuitive even when correct. More support can hurt a candidate (monotonicity) — so the count can behave in ways that are hard to explain to the voter whose ballot "did the opposite."

So the honest picture: the ballot is simple; the count needs a computer and a central tally. That's the half "look how simple it is" leaves out.

STAR is simple on both halves

The ballot — at least as easy to mark: - 0–5 is the familiar five-star mental model (every rating app you already use). (Why a score ballot differs from a rank ballot: Scores vs. Ranks — Don't Confuse Ranks and Ratings.) - You rate each candidate on their own — no forcing a complete strict order among a field of strangers ("is she my 4th or my 5th?"). - Equal scores are allowed — you can say "I like these two the same" (the repo's Equal Support). Standard IRV ballots forbid equal ranks, and many even cap how many you may rank. - One ballot carries both preference and strength — order and how much — where a rank carries only order.

The count — stays simple and summable: - Round 1: add each candidate's scores. A column sum. Precincts report totals, exactly like today. - Round 2: one automatic runoff between the top two — a single head-to-head majority.

Two transparent steps you can do on a napkin and audit on a precinct table.

See it — same winner, very different counts

→ method_comparisons/_main/cases/count_simplicity_star_vs_irv.yaml (45 voters, 5 candidates). Carmen, the broad consensus, wins under both methods — so this isn't about who should win. It's about the work the count takes.

  • STAR (STARVote_LH_tabulation_engine/): add the scores → Carmen & Andre are the top two → one runoff → Carmen. Done.
  • RCV-IRV (06_Other/RCV_IRV/RCV_IRV_tabulation_engine/): no first-round majority, so it eliminates Evan, transfers his ballots, rechecks; eliminates Dana, transfers, rechecks; now Carmen clears half — three rounds of eliminate-and-transfer for the same answer.
# run both on the one file:
python3 STARVote_LH_tabulation_engine/starvote_larry_hastings.py method_comparisons/_main/cases/count_simplicity_star_vs_irv.yaml
python3 06_Other/RCV_IRV/RCV_IRV_tabulation_engine/rcv_irv_tabulation.py    method_comparisons/_main/cases/count_simplicity_star_vs_irv.yaml

How to say it (without attacking anyone)

Don't call it a "bait and switch" — you don't need to, and it sounds like motive-hunting. Make it a precision point:

"Ranking is simple to mark — agreed, fully. But 'simple' belongs to the ballot, not the count. IRV's tabulation takes rounds of elimination, can't be added up from precinct totals, and needs software and a central count. STAR keeps the easy ballot and an addable count — and its ballot says how much you support each candidate, not just the order."

Give credit on the ballot; hold the line on the count. That's the version no one can call unfair.

Cross-references