[EM] IRV vs Plurality

[Abd ul-Rahman Lomax]

At 02:22 PM 1/10/2010, Terry Bouricius wrote:
>Stephane,
>
>Although Abd often asserts that IRV replicates FPTP results, I don't think
>he is claiming that in the last Burlington election.

That's correct. I make that claim with respect to nonpartisan 
elections, where the data is quite solid on this.

>  The plurality leader
>was the Republican Kurt Wright with 33%. He presumably would have won
>under FPTP.

That's highly speculative. There hasn't been a Republican mayor of 
Burlington in many, many years. Rather, the voting method allowed 
certain kinds of voting. We cannot predict how voters would have 
voted with Plurality, because they would have made compromises, many 
of them. Certainly a Republican voter in Burlington would have known 
that the Republican wasn't likely to win.

But it's possible. This is not the ordinary spoiler effect, though. 
It's a three-major-candidate effect, which is quite unusual in the 
U.S. for partisan elections.

>  However, as the weaker candidates were eliminated (first the
>Green and Independent, and then the Democrat) the Progressive Party
>incumbent mayor who was in second place in the initial first-choice tally
>won the runoff.

That's right. But, it must be noted, the "weaker candidate" as 
defined by IRV was preferred by a strong majority of voters over the 
Progressive who won. What happened with IRV is that it concealed the 
second-rank votes of the Republicans, officially, it never even 
counted them. There is some great video on this from SJVoter.

http://www.youtube.com/watch?v=KAGlVOi2sEY&feature=channel

I just discovered SJVoter's video work. Terry, the IRV campaign is in 
serious trouble.

>Of course, there is the additional factor that a change in voting rules
>would likely change both campaign tactics as well as fears about spoilers,
>and whether all five candidates would have even run.

That's correct. Plurality works well, of course, when there are two 
major candidates and no serious spoilers. "Serious spoilers" means 
that there are candidates who actually draw off votes from one of the 
major candidates, but not the other, votes which involve serious 
preference strength. I don't consider as a "serious spoiler," a third 
party candidate who gains votes from people who might otherwise not 
really care about the result, even if they have some preference left....

>It is certainly
>possible that the Democratic Party candidate would be dismissed as a
>"spoiler" with the Republican challenger and Progressive incumbent being
>seen as the "credible" candidates.

Its possible for that perception to occur, but not once the actual 
voted preferences are known. The spoiler, if any, was the Republican. 
Which is odd, considering that he had the largest first-preference 
support. But that is merely the result of a lack of serious spoilers, 
so to speak, on the Republican side. It's been claimed that IRV 
favors extreme results and discriminates against compromise winners. 
This is, I'm sure Terry remembers, one of the major criticisms of 
Robert's Rules of Order of IRV. Burlington demonstrated that, 
demonstrating that the claims of FairVote that such scenarios were 
rare, didn't actually occur under IRV, were false. Center squeeze 
isn't rare when there are three major candidates in a partisan election.

>  It is also quite possible that the
>Independent with around 10% first choice support would not have run if
>FPTP had been used. Voters rather universally ranked their true favorite
>choice as number one, but that wouldn't have been true under FPTP.

The candidate may have run but may have gotten much lower support. 
That's why Warren Smith writes about the "incubator effect" of Range 
voting, which effect would also apply to Bucklin. There is no reason 
at all not to vote for your favorite in first rank in Bucklin. Under 
some very unusual scenarios, maybe, adding an additional first rank 
vote *might* be optimal, if it is permitted, but this would be so 
rare that I'd not advise voters to even think about it.

I do assume that, at present, IRV or RCV voters are voting sincerely 
for first preference. However, with IRV, this can be a seriously 
suboptimal vote, I'm pretty sure that the Republicans who relied on 
the promises that they could safely vote for Wright in first rank 
are, some of them, regretting that vote, for it gave them the worst 
result, for most of them.

Under Bucklin, we would see, I'd predict, more truncation, less usage 
of additional ranks, but there would still be enough that *usually* a 
majority would be found, even with substantial numbers of candidates. 
But if candidate counts increase without limit, almost any voting 
system can be overwhelmed, unless Asset techniques are used.

>IRV resulted in a VERY different set of dynamics than would have existed
>with FPTP, so it is impossible to say with any certainty what the outcome
>would have been.

That's correct.

>  It is also noteworthy that the current debate is NOT
>about substituting typical FPTP, but rather FPTP with a 40% requirement.
>Under that scenario, it is quite likely that the Progressive would have
>won as well, since no candidate reached 40% initially, and IRV replicated
>the likely runoff outcome.

But you are assuming as "quite likely" that the Democrat would have 
been eliminated. Maybe.

I oppose a 40% requirement, it's the wrong measure. I do consider top 
two runoff to be better than IRV, but, without modification, it 
doesn't address center squeeze, which is the problem in Burlington. 
Use a better advanced method -- which can be simple as Bucklin -- and 
require a majority, or, alternatively, a vote margin above a certain 
level such that a "comeback election" is highly unlikely.

I've come, Terry, to realize that Bucklin was quite an advanced 
method, and the historical problem was partly that it was oversold as 
a method of finding majorities without runoff elections. That should 
sound familiar! No single-ballot method can guarantee majorities, but 
some are more efficient than others at finding them. IRV, with 
optional ranking, as is politically necessary in the U.S., not only 
does not find true majorities, when runoff rounds of counting are 
needed, but we can expect that to increase, based on the Australian 
experience with OPV. Bucklin, likewise, will experience higher levels 
of truncation as voters become familiar with it, but when there is a 
true spoiler effect operating, we can expect that Bucklin will either 
head it off or truncation will cause majority failure and thus a 
runoff. With Bucklin, under majority failure conditions, all the 
votes have been counted, which is quite unlike IRV, so Bucklin will 
make the best possible effort to find an actual majority, and is also 
more likely to pick the best top-two candidates.

And the requirement need not be as strict as "majority."

In a plurality election, if we have 49%A, 30% B, and 21% other, the 
likelihood of a comeback is practically zero. If this were a Bucklin 
election, and those were first preference votes, we'd have a sum of 
votes, at the end, greater than 100%, and the likelihood of there 
being enough second rank votes to take A over the majority level for 
explicit approval is very high. A runoff would be avoided even if the 
requirement is for a majority. There was an IRV election like that in 
San Francisco, perhaps you remember it, Terry. There was majority 
failure, sure, but only by a few votes.

What's needed, Terry, is for communities to start to consider voting 
systems overall, gathering information and knowledge, and making 
deeper choices than the simplified ones being presented to them by 
activists. The ideal voting method may vary from community to 
community. IRV has been sold in places where it really wasn't 
appropriate and where whatever arguments can be made for IRV actually 
didn't apply. But seemed to apply.

Multiwinner STV is a reasonable method, under some conditions. It was 
the original goal for the organization that preceded FairVote. For 
shallow political reasons, IRV became the centerpiece, but the 
defectiveness of IRV has been known for more than a century, I found 
academic criticism of it from the nineteenth century. Even the much 
better Bucklin system (called American Preferential Voting in the 
political literature from the second decade of the nineteenth 
century) was rescinded, but not because it was an inferior system; 
rather, the only cogent argument against it seems to have been that 
it was oversold as finding majorities without runoffs. That is not 
the ideal application of Bucklin. It is a method which can properly 
*reduce* the need for runoffs, and more effectively and efficiently, 
quite likely, than IRV.

But because FairVote contemptuously dismissed academic and expert 
criticism as "ivory tower," and refused to seek consensus with 
experts and more informed voting system activists, it has done great 
damage to the cause of election reform in the U.S., opportunities 
were missed and wasted and people will now be suspicious of any 
reform. Perhaps they should be suspicious, but ... reform capital has 
been spent and is likely to ultimately be wasted. We need to stop 
this, Terry, and you could be part of the solution instead of part of 
the problem, and I sense that this is possible for you. Am I wrong?