[EM] Why I think IRV isn't a serious alternative 2

[Abd ul-Rahman Lomax]

At 08:06 PM 12/24/2008, Terry Bouricius wrote:
>Abd,
>
>Abd wrote about "center squeeze":
><snip>
>The problem happens with
>reasonable frequency with Top Two Runoff, and the principles are the
>same. *In this way,* IRV simulates TTR, though, in fact, it is a
>little better in choosing among the remaining two. IRV would not have
>elected Le Pen. But it missed the very clear, and broadly supported,
>Jospin, who would have won against Chirac, and not by a small margin.
>If Jospin had gotten 1% more of the vote, the runoff would have been
>between Chirac and Jospin, the voter turnout would have been lower
>because of lower preference strength, but I'd have predicted, from
>what's known, 70% or more of the vote for Jospin.
><snip>
>
>It sounds as if you are saying IRV would have missed electing Jospin over
>Le Pen and Chirac...but perhaps by "it" you mean two round runoffs???

Sorry. Yes, "It" referred to TTR. (top two runoff). IRV would not 
have elected Le Pen, as I wrote. It *might* have elected Jospin, 
because Jospin was in third place, and before being eliminated, would 
almost certainly have gathered enough transferred votes to pass up Le 
Pen, who would have been eliminated instead. I.e., in this case, IRV 
would have gotten it right. IRV fixes *some* of the Center Squeeze 
situations that the more primitive FPTP primary in standard Top Two 
Runoff misses.

But it misses others. If we look at a close three-candidate election, 
all that has to happen is that the compromise winner, who could be 
the second choice of practically everyone, if not their first choice, 
is edged out by the other two. That's a rare circumstance in a 
two-party system, for sure, but it could be the death of a third 
party that fronted what came to be a spoiler with a vengeance.

>If
>you want to promote two-round runoffs over IRV, you shouldn't use the
>example of the French election, because it shows the exact opposite of
>what you seem to claim here.

Terry, I'm not a promoter, politician, or die-hard advocate, I don't 
pick and choose my arguments for political effect; rather I examine 
the issues. I emphasize certain points that I think important, I'll 
return to them, but I try not to be imbalanced; like Warren Smith, I 
will write stuff that appears to be contradictory to what might seem 
to be my agenda.

The French election indeed shows a failure of Top Two Runoff, but the 
failure is in the first round and in the runoff rules. If the French 
system allowed write-ins in the runoff, and the runoff were, say, 
Bucklin (two candidates on the ballot, plus a write-in is possible), 
I'd predict that the motivation there would have been sufficiently 
strong for an active write-in campaign, but, knowing the danger of a 
write-in as to spoiler effect, the write-ins for Jospin would have 
been accompanied by second-preference votes for Chirac. But even 
without those, in that case, there was no danger that Le Pen would 
have won. He got about 20% of the vote. Even if Jospin and Chirac had 
split the rest, one of them would have won.

>  That French presidential election underscores
>how IRV also is better than runoffs that reduce the field to two after the
>first count.

However, Terry, don't you realize that top-two batch elimination is 
called "IRV" when it comes to listing implementations of IRV in the 
U.S. by FairVote? Batch elimination would have shown the same effect; 
I would guess that voting patterns would have been the same, except 
that additional preferences would have been added. Jospin would have 
been eliminated, quite likely. Same problem.

The more sophisticated sequential elimination, one or only a few 
hopeless candidates at a time, is indeed better. But a Bucklin primary

>  Extremist Jean Marie Le Pen only gained the runoff with 17
>percent of the vote (barely more than his share in past elections) because
>the center-left split its 40 percent-plus pool of votes among several
>candidates. Under IRV, by reducing the field gradually, the center-left
>vote would have coalesced behind prime minister Lionel Jospin, putting him
>well ahead of Le Pen and within striking range of Jacques Chirac. This is
>an example of how IRV can avoid the center squeeze that is far more
>probable with your favored two-round runoff method.

Be more specific, Terry. Sequential elimination, unlimited round IRV, 
not just any IRV, does indeed avoid the particular Center Squeeze 
situation encountered in France. In that situation, I consider it 
almost certain that STV-IRV would have, correctly, chosen Jospin. But 
it misses others.

What do you think the American Preferential System would have done in 
France? It would also, I think it certain, have elected Jospin. With 
a whole lot less counting fuss. Bucklin and IRV will usually come up 
with the same results. Except for the 3-way Center Squeeze situation, 
where Bucklin, because it doesn't eliminate candidates, but counts 
all the votes, is more likely to get it right.

I don't "favor" two-round runoff with partisan elections in a 
multiparty environment, not with an FPTP primary, which suffers even 
more extensively from Center Squeeze than IRV.

However, two-round runoff does make better choices among the top two, 
that's clear. And we can fix the other problem by using a better 
method in the primary. You *could* use IRV, run it two-winner if 
nobody gets a majority. But better, and cheaper, use Bucklin, the 
American System.

>Another shortcoming of two-round elections is the sharply lower voter
>participation (primarily among lower income voters) typical in one of the
>rounds of a two election system.

This is alleged as a shortcoming, it's actually a benefit. In order 
to understand that, it's necessary to look at what voting system 
theorists have largely neglected, preference strength.

>  I know you have written favorably about
>such drop off in voter turnout as an effective method of "compromise"
>(voters who don't care enough stay home). I disagree.

Based on nothing, essentially, but pure assumptions having no solid 
basis in fact.

First of all, lower turnout only happens when voters don't care, they 
have no major preference between the candidates on the ballot. (Or, 
as an unusual alternative, between a write-in candidate and those on 
the ballot, see the Long Beach mayoral election where the mayor was 
re-elected even though she was disqualified from being on the ballot. 
She won a plurality as a write-in in the primary; because she was 
also ineligible to be on the runoff ballot, but still won a plurality 
against the single candidate on the ballot. Because there was another 
write-in candidacy, she narrowly missed getting a majority in the runoff.)

Turnout when voters care can be higher in a runoff than in the 
primary; that was true with that French presidential election, even 
though there was never really any danger that Le Pen would win. 
Nobody wanted to take the chance that perhaps everyone would think 
that Le Pen didn't have a chance, and it was a sure thing that the Le 
Pen supporters would turn out in full strength, and, indeed, they 
did. The runoff Le Pen vote was about a million voters higher than in 
the primary. But so was the total turnout, and Le Pen was defeated about 80-20.

In Louisiana, in the famous Lizard vs. Wizard election, runoff 
turnout also exceeded that in the primary, Louisiana voters wanted to 
make it clear that they didn't support a candidate who had been a 
Grand Wizard of the Ku Klux Klan, even though it meant voting for the 
odious Democrat who was the alternative.

However, the more usual result in a top two runoff election is that 
the general preference strength between the two candidates is 
relatively low. From the point of view of most voters, the worst 
candidates have been eliminated; for many of these voters, there is 
little or no preference between them, the worst outcome isn't a 
disaster. For other voters, they are equally useless. Consider those 
"poor voters." If the runoff somehow manages to include a true 
advocate for the poor, they will turn out in droves. But more 
commonly, both candidates represent the status quo.

Terry, there is a reason why top two runoff is associated with strong 
multiparty systems. That a Le Pen could make it to a runoff was 
useless in his case. But what if that had been a more generally 
acceptable candidate from a minor party? Suddenly that party has a 
chance to make its case to the electorate. Sometimes they win. To win 
with IRV takes a huge leap for a minor candidate, the candidate must 
jump all the way from minor status to winning, with one ballot. With 
Top Two Runoff, the gap is smaller, the candidate must only make it 
to second place.

Now, make that a little easier, and make it not miss the compromise 
candidate, so that if there is majority failure, the compromise 
candidate is in there. Bucklin probably does it.

In any case, as you know, Bucklin was quite widely used in the U.S., 
there were 52 cities, apparently, that implemented it, as of the last 
article I've found. There was huge "momentum." What happened? We 
don't really know. I don't find articles on the loss of Bucklin. That 
alone is a bit mysterious! The FairVote propaganda on it seems to be 
mere assumption. Low usage of additional ranks afflicts Optional 
Preferential Voting as well as those Alabama Bucklin primaries. My 
guess is that the same forces that stopped Ann Arbor Preferential 
Voting (IRV) are the ones that stopped the American System momentum, 
and reversed it.

Are you, in your "I disagree," claiming that voter preference 
strength has no effect on turnout? Do you think that voter 
satisfaction with outcome is unrelated to preference strength? What, 
indeed, do you consider would be the primary standard for determining 
election method performance, assuming that somehow we could read the 
minds of voters?

And is the standard different if the scale of the election is 
different? I.e., if a method of determining the wisdom of a choice 
works with, say, 10 voters, why would it not work with a thousand or 
a million voters? (Except, of course, for the complexity of gathering 
the data and the calculations.)

If voters with low preference strength exclude themselves from 
elections, *which they do*, then results will be, I contend, 
improved. Low preference strength is associated with low knowledge, 
sometimes, and sometimes with intrinsic similarity of the candidates. 
In either case, leaving the election to those who care more about the 
result, for whatever reason, with this being self-selected, almost 
certainly improves results. It's one of the reasons that Plurality 
works better than theory might predict. Those who don't care either 
don't vote in the election, or they vote more or less randomly, and 
the votes of these tend to cancel out.

Really, I'd think it should be obvious: if poor voters face 
significantly more difficulty in voting than, say, affluent ones, 
something is wrong with election access that should be fixed. Other 
things being equal, so the only obstacle is the voter time, some poor 
people may not be able to leave work, indeed. But some may be 
unemployed, balancing this or even swinging it in the other 
direction. More affluent people tend to be busier. No, I think the 
reason for poor turnout among the poor, to the extent that this might 
be true (I've not seen good data on it), would be that they don't 
perceive the alternatives as important to them, it's Tweedledum and 
Tweedledee. Give them a true advocate of the working people or the 
poor, making it into the runoff, they will turn out. It happens.

IRV, on the other hand, never gets that far. With great effort, such 
a candidate might sometimes make it up to second place, which isn't 
good enough. There is no opportunity for differential turnout to help 
this candidate. There is no opportunity to make the case to the 
people, now being taken seriously.

IRV works, basically, to stop third parties from spoiling elections 
for the majors. It doesn't allow them to actually win, in practice. 
Know of any examples? Please don't cite Burlington! The winner there 
was, in that context, a major party candidate! He'd have won under 
Plurality, almost certainly. (He won the first preference vote, 
transfers didn't change anything.)

This is the fact: IRV doesn't find majority winners, when there isn't 
a majority in the first round. The long-standing tradition in 
democracies is that a majority is based on a single election, on the 
ballots cast in that election, and that few turned out for the final 
vote is moot. If they care, if they worry that the outcome will be 
harmful, they turn out to vote. That's true in deliberative bodies, 
that's true in all kinds of elections. Top two runoff almost always, 
then, finds a majority, and trying to explain this away by claiming 
that the lower turnout -- sometimes, not always! -- in the runoff is 
such that the winner got less votes than the loser in the primary, is 
pure nonsense. Different voters, different conditions, different 
votes. We don't know, at all, what the *total* votes would be, and 
some of those voters changed their minds.

Lower preference strength means two things: lower turnout, and 
increased likelihood of a voter changing sides. It's important to 
understand both of these, and pure preference analysis completely misses it.