[EM] IRV vs Plurality

[Abd ul-Rahman Lomax]

At 09:30 AM 1/13/2010, Terry Bouricius wrote:
>Juho,
>
>That was a good summary of IRV and Condorcet dynamics, and how their
>different weaknesses might be perceived by a citizenry.  I would like to
>add one more to your list. Different voting systems provide different
>incentives for candidate behavior and campaigns and thus voter
>information.

Indeed. And, in fact, this is one reason why Plurality *often* works 
much better than voting systems theorists would expect. And why 
Top-Two runoff does likewise (and even better).

>It has been argued that IRV tends to reduce negative campaigning, or makes
>campaigns overly bland (depending on your stance), because in addition to
>seeking first choices, candidates want to reach out to the supporters of
>other candidates.

It's been argued, for sure, but it's never been shown. What happens 
is that minor candidates with no hope do cooperate, but this 
incentive actually operates in the other direction for frontrunners, it seems.

>However, with Condorcet rules, it is possible for a
>candidate to win in a crowded field while receiving no first choices at
>all.

Horrors! The candidate must really be bad, not even his or her mother 
votes for him, nor, indeed, does the candidate vote for himself or 
herself. I love these objections to voting methods that are based on 
utterly preposterous scenarios and expected knee-jerk responses to 
them. However, there is a legitimate point here, let's look at it.

>There haven't been any real-world high-stakes elections to know for
>certain what effect this might have, but it would seem reasonable to
>expect candidates to avoid taking stands on controversial issues.

No, actually, at least not more than happens at present, where 
candidates try to avoid opposing the positions of large blocks of the 
public, and will attempt to present themselves differently to 
different interest groups, whenever they think they can get away with it.

The problem is that if you make yourself as bland as possible, you 
will lose your support base, those highly motivated to turn out and 
vote for you, work for your election as a volunteer, contribute funds 
to help you gain name recognition, etc. Fatal under most realistic 
voting systems, including Range, IRV, etc.

>Candidates would have an incentive to campaign just using a vacant theme
>of "I promise to listen to YOU."

This is supposed to be new and only hypothetical? Sorry, Terry, I 
vote against candidates like that, and I think I'm not alone. I'll 
vote for a candidate whom I *actually trust* to listen to the 
constiuents, but not one who will not disclose his or her own 
position, because I can't trust the latter to vote intelligently and 
honestly. I don't want a rubber stamp in a legislature or office, I 
want someone who will not only listen, but make reasonably decent 
decisions as well, *after* having listened. Someone who won't tell me 
what they think, who avoids revealing personal positions, that's a 
very big negative for me.

Unfortunately, the present systems encourage exactly this.

>IRV seems to strike a reasonable balance between appealing for a strong
>core of supporters (the only requirement  in a plurality election with
>many candidates) and also developing broad appeal as an alternate choice.

The problem is that a political expedient is mistaken for a desirable 
quality. And it's just plain bullshit. IRV favors extremists, not 
centrists. And not *real* centrists. I'm afraid that Terry is 
reasoning backwards. He's long worked for IRV, so he is making up 
reasons why it's a good method, a "reasonable balance," even though 
anyone who has studied voting systems without this kind of activist 
bias knows that IRV performs far from reasonably.

To be clear, IRV does not favor serious extremists, but rather it is 
known to foster a two-party system, where any minor parties are mere 
appendages to a major party, and tend not to last. Terry focuses, in 
his comment, on individual candidates and how they will comport 
themselves, but probably most voters where parties are involved vote 
based on general party affiliation as as strong a factor as 
individual "promises." As has been amply explained in this excellent 
video, http://www.teach12.com/ttcx/VotingFreeLecture.aspx, when there 
are two parties, they are strongly motivated to position themselves 
near the median voter, i.e., such that their range of supporters are 
to one side or the other of the median. Candidates, in primary 
elections or other internal party process, will be motivated to 
position themselves at the median of the party, so, thinking of this 
as a linear spectrum, at 25%; but this is modified by considerations 
of electability, so the push will be up, toward the middle. Then, 
once nominated, they will attempt to present themselves as even 
closer to the median.

A third party attempting to rise up in the middle gets slaughtered 
under IRV, through center squeeze, so that's next to impossible. IRV 
is only effective, from some points of view, with respect to minor 
parties that are more extreme than the major parties.

But the whole consideration is different with nonpartisan elections, 
where IRV is apparently a fish bicycle, producing results that 
closely imitate Plurality. We can assume, with IRV, that people will 
indeed vote their first preference, unless they happen to actually 
understand the system in a center squeeze situation, and center 
squeeze is generally a partisan phenomenon, not present in 
nonpartisan elections as easily. So, true or not, most voters will 
indeed vote their first preference in IRV. But for reasons about 
which I can only speculate and infer, the plurality leader in a first 
IRV round, in a nonpartisan election, almost always wins after vote 
transfers are done, and the phenomenon is deeper than that; when I've 
looked, second position also remained second position.

IRV, in nonpartisan elections, and at substantial expense, imitates 
Plurality. IRV only makes much sense in a two-party environment where 
minor parties can only become spoilers. And I'm not sure that taking 
away the right to spoil elections is a good thing! It, at least, 
gives some clout to a minority than can command sufficient allegiance 
from voters to make a difference!

But there are much better ways to legitimately empower minority parties.

>Condorcet tips towards the broad appeal alone. Condorcet would seem to
>encourage candidates to simply avoid alienating anybody, with little need
>to develop strong core support.

It's pure speculation, neglecting the rest of the system, and 
assuming full ranking, which isn't at all likely in most real 
political applications.

>Thus, I wonder if Condorcet would "dumb down" campaigns to the point that
>voters would have even less information to evaluate candidates by.

Tell you what. Surely we should know before much more investment is 
made in building fish bicycles. How about FairVote activists start 
encouraging wider experiment with voting systems? Surely if IRV is 
the best system, it will show up this way when there are real voting 
examples and situations whcih can be compared. Eh? Modest proposal?

>A candidate who flew below the radar, such that no voters had any negative
>opinions of the person, just might win, even if finishing in last place in
>terms of first choices.

This presumes deep ranking. It presumes some advantage to "no 
negative votes" without any positive votes. And it completely 
neglects the other alternatives, only comparing full-ranked 
Condorcet, which isn't a likely option, period. Rather, ballots will 
*all* be truncated, necessarily, because, hey, what about write-in 
votes? Are these to be allowed? They aren't in Australia, with IRV, 
to my knowledge, because it would screw up the absolute majority 
requirement for mandatory full ranking, that provides a handy IRV 
vote for the voters next-to-the-bottom-of-the-barrel candidate.

>I suspect the voters wouldn't be happy, even
>though that was the logical result of their ballots.

I'm certainly not advocating a Condorcet method, even though a 
Condorcet winner is *usually* a better choice than one who would lose 
in a direct contest. Here is the reality behind Terry's biased theorizing:

IRV is replacing top-two runoff because the latter is politically 
vulnerable based on alleged higher (or actually higher) costs. 
Top-two runoff is a system most often used for nonpartisan elections, 
probably because partisan elections in a two-party system handle the 
first round in the nomination process -- or "primary" process. In 
nonpartisan elections, positive name recognition seems to be the 
primary factor, which is why IRV with such elections simulates 
Plurality, I strongly suspect. It makes no sense in that context.

For nonpartisan elections, it's become quite clear to me, using IRV 
is functionally the same as using Plurality, at higher cost. FairVote 
is selling IRV to jurisdictions that valued gaining a majority vote, 
being willing to support the cost for that, over Plurality. IRV is 
sold as a way of finding a majority without the inconvenience and 
expense of a runoff election, and, as an additional argument, it's 
claimed that low turnout in runoff elections is a Bad Thing, when, in 
fact, it pretty clearly means that the electorate is not exercised 
about the difference between the two remaining candidates.

In fact, with TTR, if the electorate thinks that both remaining 
candidates are bad, and in many jurisdictions that allow write-in 
votes in runoffs, the electorate can fix it. If the preference 
strength between the top two and a compromise candidate is strong 
enough, they can and will succeed. And using an advanced voting 
system for the runoff will prevent a spoiler effect there, and if an 
advanced method is used for the primary, it could reduce the need for runoffs.

What advanced system? Well, Robert's Rules of Order recommends 
Preferential Voting. In a massive public relations coup, because it's 
been widely copied, FairVote deceptively claimed that RRO recommends 
"Instant Runoff Voting," and this was such a "reasonable" claim that 
I bought it for quite some time before I noticed the problem. RRO 
requires a true majority; if there is no true majority, *the election 
fails and "must be repeated."* Not even a "runoff," there is no 
candidate elimination, only a more informed electorate in the next try.

But RRO also notes the center squeeze problem, in a reference to the 
single transferable vote method as failing to find "compromise 
candidates." This is a reference to center squeeze. And it also notes 
the value of additional information in repeated elections, which 
surely applies to runoff elections under TTR.

There is a much better method as a candidate for a runoff-reducer: 
Bucklin. While it technically fails the Later No Harm criterion, that 
is a singlarly weak criterion. I do not consider it a harm if my 
second preference vote elects a good compromise, nor, if I were a 
candidate, would I consider that a harm to society, and candidates 
run for society, not for themselves. We hope! The purpose of 
elections is not for some faction to become predominant, but for the 
electorate overall to be maximally satisfied with the result. Bucklin 
fits with how people are accustomed to voting, and the decision of 
each voter whether or not to add additional preferences is properly 
based on one simple consideration, if a true majority is required in 
the primary:

Would I prefer to vote in a runoff election, or would I prefer to see 
*this candidate* elected. If I'd prefer a runoff, I'll not vote for 
the candidate. If I find the candidate adequately acceptable, I'd 
vote for him or her, at an appropriate rank. Bucklin is cheap to 
canvass, so cheap that there is really little cost. It's really 
"instant runoff approval," for approval voting, used for a primary, 
would have a similar strategy I could recommend. But it allows the 
expression of first preference, and provides *limited later-no-harm 
protection*. I.e., my second-rank vote isn't counted unless it's 
clear that my first preference is not going to win based on first 
preference votes.

(The main objection to Approval voting is its inability to express a 
first preference *and* cast an effective vote if it happens that 
one's first preference isn't a frontrunner. Bucklin fixes this.)

And Bucklin worked. It was very popular. Why it was rescinded isn't 
clear in most cases, I've been unable to find much reference to it, 
this is a place where some student could do a really valuable service.

But three related reasons are known, and it should give FairVote 
activists pause.

1. It was sold as a method to find majorities in a single election. 
Familiar, FairVote activists?

2. It didn't find majorities in many elections. Particularly in 
primary nonpartisan elections to determine party candidates, that's 
the example I know.

3. In some elections, only 11% or so of voters added additional 
preferences. Supposedly, this made the use of the method an 
unnecessary complication. (This was a statistic about the party 
primaries mentioned.)

A similar argument was made by the board of the IEEE when it 
rescinded Approval Voting for its own elections. But note: most 
voters in most elections don't need to add additional preferences. 
Only those who support non-frontrunners need to do so, and these may 
be even less than 11%. It's a phony reason, obviously, not to use 
Approval, because there is no significant cost to approval. If nobody 
adds extra votes, no cost at all!

No, the IEEE board had, I'm sure, an unstated reason for dropping 
Approval. One that the membership would not like.