[EM] Better Choices for Democracy

[robert bristow-johnson]

> On 06/22/2025 6:56 PM EDT Ralph Suter via Election-Methods <election-methods at lists.electorama.com> wrote:
> 
> 
> More than a "roadmap" or strategy is needed. As impressed as I am with the arguments that RBJ, Better Choices for Democracy, and others have made in support of Condorcet methods as opposed to IRV, they aren't persuasive enough.

Thank you, Ralph.

> First, Condorcet advocates need to explain that the reason IRV was adopted many years ago in Australia and elsewhere is because it can easily be implemented with hand-counted ballots. 

Well, even for IRV, each ballot pile has to be handled N-1 times, worst case.  (N is the number of candidates.)  And this ballot pile has to be centralized (which makes it massive statewide or for a large city) or there has to be a secure communications link between all of the polling places (where the smaller piles of ballots are) and the central election authority.

With Condorcet, each ballot pile has to be handled N(N-1)/2 times.  And these ballot piles can be split up into smaller piles (a random division of piles) to distribute the job out to more people working simultaneously (because tallies can be added).  Without that 2-way secure communications link, you can't do that with IRV.  Even so, for four candidates (N=4), it's 3 handlings of the ballot pile for IRV and 6 handlings of the ballot pile for Condorcet.  It's more laborious, but just as simple as IRV.u

> Condorcet methods would have been impractical until 20 or so years ago (i.e., 5 or 10 years AFTER FairVote was formed), even for elections with relatively few voters and candidates.

Burlington Vermont was doing IRV in 2006 and we already had optical scan ballots.  The machines used were AccuVote https://verifiedvoting.org/election-system/premier-diebold-dominion-accuvote-os/ .  Besides tallying the local vote count for every office, they had removeable memory chips that had the raw data for each ballot and utility for outputting a file with simply the discrete matrix of ovals that were adjudged to be marked or filled in.  Then there was another company (or maybe a consortium of IRV activists/professionals, similar to RCVRC is today) called Voting Solutions that had software ChoicePlus Pro that took that matrix of ovals and a mapping of that to the IRV ranked ballot, and from that it performed the IRV rounds, tallying active votes, eliminating candidates, and transferring votes.  This existed at the turn of the century, likely before because Cambridge MA has been doing STV for decades.

Condorcet would have just as practical then as was IRV, ever since electronic vote tallying became practical, even for small towns.

> The reason they are practical now is because computerized voting machines and scannable ballots (which can easily be audited with small amounts of hand counting) have made them so.

We've had optical-scan ballots in Vermont since, at latest, 1990.  Computers have existed and been used for tallying votes in elections since, I dunno, 1960.


> Second, now that Condorcet methods can much more easily be implemented, they have several important advantages over IRV:
> - They are "precinct summable," which simply means that unlike IRV, there's often no need to wait till all votes are counted before an almost certain winner can be determined.

It's not just the wait (but it *is also* about the wait), but it's about the process transparency and the redundancy (double-checking election results) that is lost when we move from FPTP to IRV.  But we don't lose it with Condorcet (or Bucklin or Borda) or Score/STAR or Approval.  It's just IRV that loses it.

Remember that Summability is what exposed the Venezuelan presidential election (about a year ago) as stolen.  Now, imagine Venezuela having a judicial branch that was both independent and had power that the executive could not compromise, then Summability (and the power of the courts) would have protected that election from being stolen.

I just find it incredulous that Americans (and others in the West) are so easily persuaded to sacrifice that component of government accountability.  We shouldn't put up with it, at all.

> It's rarely necessary to wait days (sometimes even weeks) to be pretty certain of the winner, as it sometimes is with IRV. This also addresses a complaint commonly made by IRV opponents (mostly conservatives) who have opposed IRV and even sponsored legislation in many states to make IRV illegal.
> 

This opaque centralization of ballots is onerous.  Process transparency in elections should be inviolate in an authentic democracy.

> - This advantage also makes Condorcet methods much more useful for polling. It would be much easier to do accurate polling of candidates in Condorcet elections than it is to poll for candidates in IRV elections.

That's because it's easier to count votes, conceptually, with Condorcet.  It's more laborious, but it's simpler in concept.

> Using Condorcet methods would also make it possible to do accurate polling of three or more possible policy options (e.g., three or more different proposed ways to to improve on the current Electoral College system for electing US presidents).

I really think that this Condorcet vs. IRV debate doesn't affect the current constitutional method that the American president is elected.  If this were to change and the Constitution was amended to elect the president by popular vote, and even amended further to elect the president using RCV, *then* this becomes relevant in the Condorcet vs. IRV debate because no one wants to centralized data from 150 million ballots.  That's when it *really* has to be Summable so that each state can run their own elections and publish official tallies that *everyone* can just add up and know who wins.

> In addition, Condorcet methods would make it possible to have referendums with three or more options (e.g.: 1) elect presidents by direct popular vote, 2) keep the electoral college but reform it to end the spoiler problem and make elections fairer for voters in all states, 3) have the House of Representatives elect presidents and enable it to replace a president with a vote of confidence, as was recently proposed by law professor Max Stearns in his book "Parliamentary America"). Using IRV for referendums would not be a good idea.
> 
> Third, better ways in addition to the above are needed to explain Condorcet voting and its advantages over IRV. One long used rhetorical device is to describe Condorcet methods (a term that makes no intuitive sense to people who aren't familiar with the history of voting methods) as forms of "Instant Round Robin Voting" or IRRV and explaining that IRRV methods simply involve examining ranked ballots to determine how each candidate in an election would fare in one-to-one contests with each other candidate, which is hard to do with hand counting but easy to do with computerized voting machines or scannable ballots.

It's *laborious*, more laborious than IRV for 4 or more candidates.  But it's simple to do with hand counting.  Just more labor, but the labor is easily and safely divided.

> Few if any people would disagree that if one candidate would defeat every other candidate in one to one contests, as would almost always be the case, that candidate should be declared the winner.

Of course, FairVote and RankTheVote and all of the satellite organizations like CalRCV or FairVote Washington, or VoterChoiceMA disagree.  The say that sometimes it should *not* be "that candidate" declared the winner.

> You can then explain that in the rare instances where there isn't a candidate who defeats all the others, there are usually persuasive ways to decide which candidate is the most preferred overall and should be declared the winner.

Yes, it has always been my approach to be completely transparent about the possibility (and factual occurrence) of cycles.  Policy makers need to be made aware of Arrow/Gibbard/Satterthwaite.  Not as experts, but they should understand that it's possible that voters can collectively vote in such a way that it's impossible to avoid a spoiled election and that some group is going to be screwed.  I've always been open to discuss that sorta unsolved problem.  However a legislative decision can be made to how to resolve a preference cycle.  And it doesn't happen in 99.6% of the RCV elections.

> Better Choices for Democracy uses the term "consensus choice voting" instead of IRRV to describe Condorcet methods.

Yes, at that meeting I was at, there was much discussion over the name (instead of "Condorcet").  I am not sure it's the best name.  I am also unhappy that they are apparently ceding the term "RCV" to Hare.  They're not saying that Consensus Choice Voting is also Ranked-Choice Voting even though the ballot is exactly the same and the meaning of ranking is exactly the same.  It's just a different way to count the ballots and determine who the winner is.

I got some issues with the decisions they have made.  I expressed my concerns in 2023 to these folks that eventually are Better Choices for Democracy.  I think it's still too Ivory Tower.  I think there's stuff they could learn from the STAR voting site.  They need to discuss, specifically what was wrong with the outcomes of Burlington 2009 and Alaska 2022 (August) and make it relatable to pedestrians.

> Only experimentation with different terms and explanatory methods will determine which terms and methods are most effective. Maybe there are alternatives to both IRRV and consensus choice voting that would be more effective than either.
> If and when Condorcet advocates really get their act together, as I hope they soon do (BCD is at best a good start), I think it's very likely that they will decisively win out over IRV advocates in the long run if not in the very near future.

This is what I meant by saying history is on our side, not so much with FairVote or RankTheVote.

> I don't mean that as a put-down of IRV advocates. I agree with Michael Garman that IRV is an improvement over plurality and that if "perfect" Condorcet methods were unachievable, less than perfect IRV would still be better than plurality and worth adopting for that reason alone.

It entrenches the wrong method in even deeper.  If RCV were ever to be widely adopted, then these avoidable errors would be happening every year, instead of once per decade.  Then, when there's public outcry, they'll say it's "the standard method used for centuries and just cannot be changed."

The time to make course corrections in a voyage, is early in the voyage.  Much less costly than far later in the voyage and the IRV advocates would just tell us that this uncorrected course result (that takes us to a different place than was promised) is where we wanted to go after all.  It's so disingenuous.

It surely appears like, Ralph, you've been persuaded.

--

r b-j . _ . _ . _ . _ rbj at audioimagination.com

"Imagination is more important than knowledge."

.
.
.