Plurality, approval, or ranked: picking a voting method for your group

Most groups never choose a voting method. They inherit one — usually “everybody name your favorite, most names wins” — and then argue about the result without noticing that the counting rule, not the group, produced it.

That is worth knowing because the choice of rule is not cosmetic. Below is a single set of preferences that produces three different winners under three perfectly reasonable methods.

The example

Eight friends, four restaurants. Everyone honestly ranks all four, best to worst:

Nobody is being strategic and nobody is lying. Now watch what the counting does.

Plurality: everyone names one favorite

Count first choices only. Ramen gets Ana, Ben, and Cara — three. Tapas gets Dev and Eve — two. Curry gets Finn and Gia — two. Pizza gets Hal — one.

Ramen wins with three votes out of eight. Five of the eight ranked it dead last.

This is plurality’s signature failure and it is extremely common in small groups. It rewards a committed bloc over a broadly acceptable option, and it gets worse as you add options: four similar choices split the vote of everyone who likes that kind of thing, handing the win to whichever distinct option has the most concentrated support. The pizza camp and the curry camp are, in effect, spoilers for each other.

Plurality is fine when there are exactly two options, or when you expect a genuine favorite to be obvious. With four options and eight people it is close to a coin flip dressed up as a decision.

Approval: tick everything you are happy with

Different question — not which do you prefer but which would you be glad to go to. Each person ticks as many as they like. Given the rankings above, a natural set of approvals looks like this:

Pizza wins with seven of eight.It was almost nobody’s favorite — only Hal put it first — and it is the option nearly everyone is content with.

Approval measures breadth rather than intensity, which is usually what you actually want from a group decision. Its weakness is the mirror image: it cannot tell the difference between “I would love this” and “fine, I suppose”, so an option that everyone mildly tolerates can beat one that half the group would be genuinely excited about. It is also mildly gameable — someone who ticks only their favorite gets more leverage than someone who ticks three — though in a group of friends who can see each other next week, that tends to sort itself out.

For scheduling, approval is not just a good method, it is the correct one. Availability genuinely is binary: you can make Thursday or you cannot, and ranking your free evenings by enthusiasm answers a question nobody asked.

Ranked choice: eliminate the weakest, redistribute

Take the full rankings and run instant-runoff. Pizza has the fewest first choices (Hal alone), so Pizza is eliminated and Hal’s vote moves to his next surviving option, Curry. Standings become Ramen three, Curry three, Tapas two.

Tapas is now lowest and drops out. Dev’s next choice is Curry; Eve’s next is Pizza, which is gone, so hers moves along to Curry as well. Final round: Curry five, Ramen three.

Curry wins. Same eight people, same honest preferences, third different answer.

Ranked choice fixes plurality’s spoiler problem and does capture some intensity. It costs you two things: filling in a full ranking is real work, which depresses response rates, and the outcome depends on elimination order in ways that are hard to explain to someone who is annoyed about the result. Notice that Pizza — the option seven of eight were happy with — was eliminated in the very first round, precisely because it was widely liked rather than intensely loved.

Which one was “right”?

There is a reasonable tiebreaker here. Check every option head-to-head against every other one. Pizza beats Ramen five to three, beats Tapas five to three, and beats Curry six to two. It wins every pairwise contest, which makes it the Condorcet winner — the option a majority prefers to each alternative individually.

Approval found it. Plurality and instant-runoff both missed it, in opposite directions. A Borda count, where a first place is worth three points, a second two, and a third one, also lands on Pizza with fifteen points to Tapas’s thirteen, Curry’s eleven, and Ramen’s nine.

The lesson is not that approval voting is always correct — there is a proof, Arrow’s theorem, that no ranked method can satisfy every property you would want at once. The lesson is that for the ordinary case of a small group picking among several acceptable options, the plurality rule everyone defaults to is the one most likely to produce a winner most people did not want.

Constraints are not votes

One thing no counting method handles: hard constraints. If someone cannot eat at the sushi place, or genuinely cannot afford the $250 option, that is not a preference to be outvoted. It is a filter to apply before the ballot exists.

Ask about constraints first and separately — dietary needs, budget ceiling, accessibility, who has a car — then build the option list out of things that survive them. Putting an unworkable option on a ballot and letting the group vote it in produces the worst outcome available: a decision that is both legitimate and impossible.

Practical guidance

• Scheduling a date or time: approval. Tick every slot you can make. The winner is the widest overlap, which is the only thing you were looking for.
• Choosing between a handful of acceptable options— restaurant, movie, activity: approval again, optionally with a “first choice” marker to break ties by enthusiasm.
• A big, expensive, one-shot decisionwhere intensity really matters — the destination for a week-long trip: ranked choice, or approval plus an explicit round of discussion. Accept that it costs more of everyone’s attention.
• Two options: plurality. It is the same as everything else at that point.

Announce the rule before you open the vote

Whichever method you pick, say what it is in the invitation, along with how ties break and when voting closes. “Tick everything that works; most ticks wins; if two tie I’ll take the earlier date; closes Thursday 9pm” is a complete constitution in one line.

Announcing the rule afterwards, or changing it once the results are visible, is the fastest way to make a group stop trusting your polls — and a rule chosen after the fact is indistinguishable from picking the winner you wanted.

One last thing worth keeping in proportion: turnout beats method. A well-chosen rule applied to four responses out of nine is worse than plurality applied to all nine. Get everyone to answer first; optimize the counting second.