[EM] inaccurate Fargo approval voting results

[robert bristow-johnson]

> On 06/09/2024 1:49 PM EDT Closed Limelike Curves <closed.limelike.curves at gmail.com> wrote:
> 
> 
> Michael—the issue is it fails much more than 1% of the time, it's just hidden from sight. We can see the same mistake in FPP. If it looks like 97% of FPP elections "got the right result", there's clearly something fishy going on in that number.
> 

I'm just saying that approx 15 RCV elections (outa ca. 500 in the U.S.) had a dramatic come-from-behind victory where the plurality leader (of just first-choice votes) in the semi-final round was defeated in the IRV final round.  That's about 3%.

> And there is! The issue is that those 97% and 99% figures are based on the incorrect assumption that which candidates choose to run isn't affected by the electoral system.

It may be.  But we don't know for sure that the assumption is incorrect either.  It could be that for most or nearly all of these elections, the same set of candidates would have run.

We try to anticipate when voters vote insincerely and it's easier to do in the case of FPTP when there are more than two significant candidates and the margin of victory was far less than the vote that the third-place candidate received.  We can, sometimes accurately, speculate that the election was spoiled and would have turned out differently if that third-place candidate had not run.

But we don't really know for sure if we didn't have ranked ballots that tell us *explicitly* who the voters for the third-place candidate would have voted for in the contingency that this candidate had not run.  But we *do* know if we have ranked ballots.

> If everyone who wanted to run for office did run, pathologies would be much more common. We saw this in Alaska, where the top-4 format made the pathologies that usually get swept under the rug by the nomination stage visible. In the very first major election with multiple significant candidates, we saw a simultaneous monotonicity/participation/majority failure.

Do you mean Alaska August 2022?  Or Burlington 2009?

> This lines up much more closely with what models predicted: in competitive 3-candidate elections, all of these pathologies should be quite common, happening around 5-15% of the time.
> 

But we just don't know.  I wouldn't necessarily be using someone's parochial model to justify changing the law in public elections.

> The reason IRV looks like it "basically works" is because its pathologies mean moderate and third-party candidates know they have no hope of winning, so they never run in the first place.

We just don't know that.  It's speculative.  It's like insisting that we *know* some FPTP election was spoiled when we just don't know for sure.  Perhaps, in Florida 2000, all those Nader voters would have stayed home if Nader hadn't run.  Or perhaps they would have leaned toward Bush (not likely, I'll admit) and the outcome would have not changes.

> At that point, in a 2-party system, all voting systems will return the same results (because it's just a simple majority vote).
> 
> I think this bears repeating: a low rate of empirical pathologies is often a negative, not positive, indicator. If your dataset has no examples of center-squeeze, that means your system is so bad at electing Condorcet winners that moderate candidates are refusing to run in the first place. 

We just don't know that.  It's speculative.

> Similarly, we'll know Condorcet methods are working if (sincere) Condorcet cycles start popping up all the time.

We just don't know that.  It's speculative.

Admittedly, I would expect cycles to be more common if Condorcet were used, but it's just a speculative expectation.  Perhaps RCV voters would rank the the candidates no differently than they had with IRV.  With the exception of a few equal-rankings, to assume that they would mark their ranked ballots differently under Condorcet than they do under Hare is speculative.  I would rather assume that voters mark their ballots with they're sincere preferences.  But if IRV starts to more often burn voters for marking their ballots sincerely, I *would* expect voter behavior to change a little.  But we just do not know for sure.

> That's how we'll know we've successfully depolarized our politics and broken free of the old one-dimensional political spectrum (where the median voter theorem protects us from cycles).
> 

I dunno.

--

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

"Imagination is more important than knowledge."

.
.
.