Election Methods — a Practical Guide

Every option described here is available in the Contest Rules panel when you set up a contest. This page explains what each method does, how the methods differ, and which one fits which situation. The examples use small numbers so you can check the arithmetic yourself.

Choosing a method — the short version

Everything else on the Contest Rules panel — quotas, thresholds, tie-breaking, skipped ranks — is a refinement of one of these.

Ranked-choice basics

In a ranked-choice election, voters rank candidates instead of picking just one: 1st choice, 2nd choice, and so on. Counting proceeds in rounds. In each round every ballot counts for its highest-ranked candidate who is still in the race. If no one has enough votes to win, the weakest candidate is eliminated and their ballots move to each ballot's next choice. A ballot with no remaining choices becomes inactive (you'll see these counted in every results table).

The point of the whole exercise: your vote is never wasted on a hopeless candidate — it moves on and keeps working for you.

RCV Lab adds one unusual freedom: a voter may rank several candidates at the same position — see Equal rankings next, or the full explanation. It is the feature that most distinguishes this site from other ranked-choice tools.

Equal rankings

RCV Lab lets a voter place several candidates at the same rank, so nobody has to invent an order they don't actually have. The ballot's weight is split equally among the candidates sharing a rank, and as those candidates are eliminated the share consolidates onto the ones still standing — the ballot only moves down to your next rank when everyone at the current rank is out. Splitting a vote therefore never weakens it.

It is off unless you turn it on; see Equal Ranking of Candidates under Ballot-handling rules for the setting and what each choice does to the count.

Full explanation, with a worked example →

IRV (Instant Runoff, single winner)

The standard single-winner form of ranked choice, used in Maine, Alaska, New York City primaries, and many cities.

• Winning threshold: a majority of the votes still active.
• Each round: if a candidate meets the threshold, they win. Otherwise the last-place candidate is eliminated and their ballots transfer to their next choices.

Example. 9 ballots, three candidates: Apple 4, Banana 3, Cherry 2. Nobody has a majority (5). Cherry is eliminated; suppose both Cherry ballots ranked Banana next. Now it's Apple 4, Banana 5 — Banana wins with a majority, even though Apple led on first choices. That's IRV doing its job: Banana was acceptable to a majority, Apple wasn't.

Try it live: the 2018 Maine Governor Primary and 2017 Minneapolis Mayor entries in the Analysis page's sample library are real IRV elections.

STV — multi-winner, proportional

Single Transferable Vote fills several seats at once so that groups of like-minded voters win seats roughly in proportion to their numbers. Two new ideas appear:

The quota. Instead of a majority, a candidate wins by reaching a quota — see Droop vs Hare below.

Surplus transfers. If a candidate receives more than the quota, the excess isn't wasted: it transfers onward, at reduced weight, to the ballots' next choices. If Meg needs 100 votes and gets 150, each of her 150 ballots keeps ⅔ of its weight with Meg and passes ⅓ on. This is what makes STV proportional — a very popular candidate's supporters still influence who fills the other seats.

RCV Lab offers two STV counting rules, which differ in how exactly they handle those transfers:

STV with WIGM

The Weighted Inclusive Gregory Method — the fractional-transfer rule implemented by the certified RCTab tabulator and used in modern U.S. multi-winner RCV elections such as the Minneapolis Park Board. When a candidate passes the quota, every one of their ballots transfers onward at reduced weight, once; from then on the elected candidate is set aside, and later ballots that would have reached them skip past to the next choice. Simple to state, well-tested, and the choice when you want results comparable with official U.S. tabulations. (Contrast with the older whole-ballot approach still used in Cambridge, MA, where a sample of surplus ballots is physically transferred at full weight — a method RCV Lab does not implement.)

STV with Meek's Method

Meek's method keeps every elected candidate "in the count" with a keep factor — the fraction of each arriving vote they retain, tuned so they hold exactly the quota. Later transfers flow through elected candidates rather than skipping them, and as ballots become inactive the quota itself is recomputed downward. The arithmetic requires a computer (that's literally why it was invented in 1969), but it distributes surpluses more evenly, especially in elections with many transfers or many inactive ballots. Used for local-body elections in New Zealand.

WIGM or Meek? They usually elect the same candidates. Choose WIGM for comparability with U.S. official counts; choose Meek if you want the most equitable surplus handling and don't need to match an external tabulator. (The Analysis page's Split View is a nice way to see them side by side on the same ballots.)

Try it live: the 2013 Minneapolis Park Board sample is a real 3-seat contest.

Quotas: Droop vs Hare

The quota is the vote count that wins a seat, chosen on the Contest Rules panel for STV contests.

• Droop Quota — votes ÷ (seats + 1), plus a margin. With 100 votes and 3 seats: 26. It is the smallest quota such that no more candidates than seats can reach it. This is the standard choice for STV, and RCTab's default.
• Hare Quota — votes ÷ seats. With 100 votes and 3 seats: 33⅓. A larger quota: winners must gather more support, surpluses are smaller, and the last seats are more often decided by elimination than by reaching quota. Historically important, occasionally preferred for its simplicity.
• Custom — any fraction between the two, if you have a specific rule to reproduce.

The sample library includes the Minneapolis Park Board election under both quotas — load each into a Split View pane to see exactly where they diverge.

Two refinements apply to either quota:

• Non-integer threshold — normally the quota is a whole number of ballots: the division is rounded down and a full vote is added (10 votes, 3 seats: floor(2.5) + 1 = 3). With this option the quota is the exact fraction plus a hair (2.5001), so a candidate wins by any margin above the true proportion. Matters mostly for small elections and split-weight ballots.
• Fixed threshold — normally the quota is recomputed as ballots become inactive (fewer active votes → smaller quota). Checking the Fixed Threshold box freezes it at its first-round value instead.

Bottoms Up

No quota at all: keep eliminating the lowest candidate until either N candidates remain (they're the winners) or every remaining candidate holds at least X% of the votes — you pick the termination criterion. Because there are no surplus transfers, it is not proportional in the STV sense, but it's easy to explain and useful for narrowing a large field (some party primaries use it exactly that way).

Single-round methods

For comparison — and for quick polls — RCV Lab also counts three non-ranked methods:

• Plurality — each voter marks one candidate; most votes wins. The baseline everyone knows, with the weakness ranked methods fix: a crowded field can elect someone most voters opposed.
• Approval — each voter marks as many candidates as they approve of; each mark counts as a full vote; most approvals wins.
• Approval (split vote) — like Approval, but each ballot's marks share a single vote (approve three candidates, each gets ⅓). Every voter contributes exactly one vote of total weight.
• Cumulative (point allocation) — each voter distributes a fixed budget of points among the candidates, doubling up if they like. The grid on the ballot becomes a point-allocation grid.

Ballot-handling rules

These options decide how imperfect or unusual ballots are counted — whether they arrive in an imported ballot file or through RCV Lab's own voting screens. (The screens prevent co-ranking when it's not allowed, but voters can always skip ranks; the skipped-ranks rule applies whenq the ballots are counted.)

Equal Ranking of Candidates

What happens when a ballot puts two or more candidates at the same rank:

• Allowed — votes split equally between them. RCV Lab's native behavior and a deliberate feature, explained in full under Equal rankings: the ballot's weight splits across the candidates sharing a rank, and consolidates onto the survivors as they are eliminated. The voting screens permit co-ranking.
• Not allowed — that rank is skipped, counting continues. A shared rank (an "overvote") in a ballot file is passed over and the next rank is read.
• Not allowed — the ballot ends there. The overvoted rank and everything after it are discarded; choices before it still count.

Skipped ranks

A ballot that marks a 1st and a 3rd choice but no 2nd has skipped a rank. Skipped Ranks Allowed sets how many consecutive skipped ranks are forgiven before the ballot stops counting — the count continues with the next ranked choice while within the limit. Enter a number, or leave it as "unlimited" to ignore gaps entirely (RCV Lab's default). A limit of 0 means any gap ends the ballot; choices before the gap still count. This matches the certified RCTab rule of the same name — Maine, for instance, uses a limit of 1.

Deciding ties

When two candidates are exactly tied for elimination (or for a seat), the Tiebreak Mode decides:

• Use Candidate Order — the candidate listed later on the ballot loses the tie. Deterministic and transparent; fine when the order is itself fair (alphabetical, or drawn by lot beforehand).
• Geometric Borda Count — the tie is broken by overall ballot support: every ranking a candidate received contributes a weighted score (1st choices count most, each later rank geometrically less). The tied candidate with less total support loses. Breaks ties by "what do the ballots say" rather than by list position.
• Random — a random draw with a recorded seed, so anyone can reproduce the result. The seed is set on the rules panel.

Whenever a tie is actually broken during a count, RCV Lab records it: the results table marks the decisive cell with a dagger (†) and lists each tie-break under the table — who was tied, at what total, and how it was resolved.

Stopping rules (Termination Mode)

For IRV and STV you can choose when the count stops:

• Stop When Winners Known — end as soon as the winners are determined. The default, and the shortest table.
• Continue Until Only Winners Remain — keep eliminating so the final round shows just the winners. Some jurisdictions publish results this way.
• Converge All Winners to Threshold — after the seats are filled, redistribute remaining surpluses so every winner's final total lands at the threshold. Emphasizes the proportional ideal: each winner ends holding exactly one quota.
• Full Elimination Then Converge — both of the above.

The stopping rule changes how the story reads, not who wins.

Want to experiment? The Analysis page lets you load any of these sample elections — or your own ballot data — change one rule at a time, and re-tabulate. Split View shows two rule-sets side by side on the same ballots.