site banner

Small-Scale Question Sunday for June 16, 2024

Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?

This is your opportunity to ask questions. No question too simple or too silly.

Culture war topics are accepted, and proposals for a better intro post are appreciated.

2
Jump in the discussion.

No email address required.

Ranked Choice Voting describes a kind of ballot design where candidates are preferentially ranked, Instant-Runoff Voting describes one way to decide a winner from such ballots.

Australia's had RCV/IRV for federal elections since 1918, and voting in federal elections has been compulsory since 1925, so it's been around a while. There's definitely better ways to decide winners than IRV from a theoretic, Bayesian regret perspective, but what's often missed about the historical popularity of IRV (and particularly its enduring path-contingency in Australia) is that it's dead simple to administer and tally compared to otherwise better methods such as approval/range. IRV solves (or solved) many of the complexities handling preferential voting in simultaneous counts across multiple voting locations in an auditable, non-destructive way (which remains one of the advantages of FPTP, for that matter). Relatedly, Australia's always voted and counted by hand (no voting/counting machines) and has generally done so quite efficiently.

Ranked Choice Voting describes a kind of ballot design where candidates are preferentially ranked, Instant-Runoff Voting describes one way to decide a winner from such ballots.

True, but the only vaguely effective lobbying/activist group for alternative voting methods in the US is FairVote and they are strongly against any voting method other than IRV and one of the ways they actively try to confuse the issue is by using term "Ranked Choice Voting" to refer to Instant-Runoff Voting, as you can see on their website.


it's dead simple to administer and tally compared to otherwise better methods such as approval/range.

I'm confused: IRV is notable for being one of the only voting methods that fails the summability criterion making it by far the hardest to tally. Does this issue just not appear in practice because there just aren't ever that many candidates so the factorial of the number of candidates stays manageable? Or is it handled some other way?

IRV is simpler to tally and audit in low-tech scenarios because the votes themselves are the physical record of the count, an advantage it shares with FPTP. In FPTP, you sort and bundle votes into e.g rubber bands of 20 and boxes of 1000. You can easily verify a count by checking that a box indeed contains 1000 votes, and they've been sorted appropriately. It's easy to update your count report just by seeing you have X boxes and Y bands.

Approval voting, Score, Borda etc require you to maintain a store of the counts independent of the physical ballots, which introduces more room for human error and complicates recounts/verification. You need to increment up to N counts for a race with N candidates, and even approval voting has 2^N-1 ballot variations that complicates sorting.

If you're thinking of terms of low-tech boxes and counting processes, IRV is an intuitive extension of FPTP because you're just opening up eliminated boxes and resorting them. Practically, it's rare for this process to go particularly deep or be particularly sensitive, and the count of votes rarely exceeds twice the votes cast.

I think we're now able to have do much better than IRV and I think there are potentially clever ways you could do tear-off perforated ballots to make the counts under approval/range more reliable, but a lot of (sometimes conspiratorial) questions about the popularity of IRV miss that it was an intuitive and practical solution at the time. If Australian preschoolers can vote on schoolyard activities with IRV, I'm sure American adults can manage.

Oh, counting votes by sorting ballots never even occurred to me. American elections almost always have several races on the ballot, so that's not really a feasible way to organize the counting. The counting is almost always done by machine anyway, with hand-counting only for recounts.

Does this issue just not appear in practice because there just aren't ever that many candidates so the factorial of the number of candidates stays manageable?

Not sure about Australia, but the UK traditionally transported all of the ballots in one race to a central location and mixed them before counting them in order to obfuscate the precinct-level results. (This is logistically trivial if the largest race is a 65,000 elector Westminster seat). So summability is not relevant. For the London Mayoralty (which used an IRV-like system where voters are only allowed 2 preferences) running the whole count at one site is a logistical headache - when the Tories changed the counting system to FPTP this allowed the counters to count the mayoral vote at borough level and sum.

The largest single IRV/AV race is the Irish presidency. Does anyone know how that is counted?