[EM] Does IRV elect "majority winners?"

[Abd ul-Rahman Lomax]

At 07:14 AM 12/31/2008, Markus Schulze wrote:
>Hallo,
>
>usually, the term "majority winner" refers to
>a candidate who is strictly preferred to every
>other candidate by a majority of the voters.

That's a "majority preference." The term is not used that way in 
parliamentary or political usage.

>However, IRV supporters usually use the term
>"majority winner" for a candidate A who can
>win a majority (or at least half of the votes)
>in a runoff between candidate A and some other
>candidates.
>
>Question: Who can win a majority (or at least
>half of the votes) in a runoff between himself
>and some other candidates? Answer: Everybody
>but a Condorcet loser.
>
>So when IRV supporters say that IRV "always
>elects a majority winner" then this is
>EXACTLY the same as saying that IRV "never
>elects a Condorcet loser".

Well, almost exactly. What is meant by the statement is a tautology. 
However you get to the last stage (eliminations? lottery? coin 
tosses?), if you only look at the votes showing for the two remaining 
candidates, it will either be a tie or there will be one of them with 
more votes. Thus, technically, the winner has a "majority of votes in 
the last round," but this is so far from how people use the phrase 
"majority of the votes" -- referring to the whole election -- that it 
is highly misleading. That's been my point. It's deceptive, 
particularly in the context of replacing top two runoff -- which has 
a true majority requirement -- a majority of the legal ballots must 
contain a vote for the winner, or there is no winner at that stage -- 
with IRV. "The winner will still be required to get a majority of the 
votes," will then be interpreted by a reasonable reader as meaning 
that the *same* requirement will be continued, only by allowing vote 
substitutions it becomes possible to find a majority -- the same kind 
of majority -- without resorting to a runoff.

Problem is, it is possible, sure, but by no means is it guaranteed -- 
and, in fact, it seems, when IRV, in nonpartisan elections, where 
there is not a majority in first preference, IRV -- three rank RCV in 
any case -- never finds one, most of the time. Bucklin is more likely 
to do this, I suspect, but not a *lot* more likely.

FairVote wanted to promote IRV in the U.S. to pave the way for 
proportional representation using STV. They knew that the complexity 
of the method would be an obstacle. Aha! Single winner STV is "like" 
runoff voting.

And so they were off. Sell IRV on the basis of being cheaper and more 
convenient, and whatever other arguments could be made up. Sell it as 
"modern" -- which is pure nonsense. It was used in the U.S. As far as 
I know, it may have been used before in San Francisco, in 1916 some 
kind of preferential voting apparently was; I have not yet found what 
kind, it could have been Bucklin.

Given that the context was top two runoff as the existing method, and 
because the goal of top two runoff is to find a majority winner (in 
the primary), they needed to be able to make the same claim for IRV. 
So they made it. Never mind that it was utterly deceptive.

It's possible to argue that IRV is a better method than top two 
runoff, but that is a quite different issue. In the discussions of 
the "Ware method" that I'm seeing, back in the second decade of the 
twentieth century, the term "majority" was used with reasonable 
accuracy. However, even there, there is some ambiguity. The last 
round winner for the Ware method, where a "majority" was not found, 
is the one with the "most votes." Not called a majority. Yet in 
earlier rounds, they *are* searching for a "majority."

This *could* be interpreted as "majority" -- true majority! -- being 
used as the standard, canvassing continues until a majority is found 
or there are only two candidates left, in which case the one with the 
most votes wins. That's the language....

Note that this was lost in recent implementations. San Francisco 
stops canvassing when there is a *round* majority, even if a true 
majority has not been found. There was one election there where the 
winner didn't have a majority, but *probably* if they had continued 
to count, the winner would have had one. *But they didn't care!* The 
winner had a majority of the votes in the round, that was considered enough.

If course, if IRV were used as a primary stage, canvassing would be 
continued to the end, if a majority hasn't been found. But I don't 
see any reason to use IRV for a runoff system primary; Bucklin counts 
all the votes, IRV does not.

By the way, one article I'm reading claims that "the Bucklin system" 
was first proposed by Condorcet in 1793, and that "it is said it was 
used for a time in Geneva, Switzerland."

http://books.google.com/books?id=LJaetzppfp0C&pg=PA3&vq=%22preferential+voting%22&dq=bulletins+for+the+constitutional+convention+1917&source=gbs_search_s&cad=0#PPA309,M1

p. 309.

>Question: So why don't IRV supporters just
>say that IRV "never elects a Condorcet loser"?
>Answer: IRV supporters don't want IRV to
>be judged by its properties but by its own
>underlying heuristic. We all know that every
>election method is the best possible election
>method when judged by its own underlying
>heuristic. If IRV supporters just said that
>IRV "never elects a Condorcet loser", then
>this argument could also be used by the
>supporters of other election methods that
>satisfy the Condorcet loser criterion.

They don't say that because their arguments are mostly not designed 
to appeal to those who have some understanding of voting systems. 
They are designed to appeal to those with little knowledge, and for 
this audience, "Condorcet" generally means nothing.

"Majority" means something to these people. Unfortunately, what it 
means to them and what it means to the FairVote cover story are 
different. (The "cover story" is what is said when someone who knows 
what "majority" means and who knows that IRV doesn't find a 
"majority," necessarily, questions the usage. "Oh, this refers to the 
majority of votes in the last round. After all, the voters have 
declined to participate in this 'runoff' so only those who did count. 
Just like runoff voting! See?"

"But... "

"Oh, maybe more voters voted in the IRV election, voting for the 
winner, than would have voted for the winner in the runoff election. 
See, this is more fair."

"But that's not what you said: You said that 'the winner will still 
be required to get a majority of the votes."

"Silly you! Do you expect us to give you all the details? We meant, 
of course, the majority of the votes among those who voted for the 
Republican or the Democrat. That's all that matters, anyway, right? 
Majority of the 'votes' means 'majority of the important votes,' 
perfectly normal usage. And, besides, Robert's Rules of Order 
considers that abstentions don't count in the basis for majority, and 
those who didn't vote for a Republican or Democrat have effectively 
abstained, right?"

This is what we get when those who have some knowledge use it to 
deceive. The ends justify the means. And anyone who disagrees with us 
is obviously a stooge of the status quo. Go away, don't interfere 
with our noble cause, all you are is wreckers, anyway, what have you 
done for voting reform?

I can say what I have *not* done. I have not wrecked and caused the 
abandonment of the best voting reform in the United States, the 
closest we have to an advanced voting system, primitive as it is: top 
two runoff.

In nonpartisan elections, it is becoming clear, it is *better* than 
IRV; in that context, IRV apparently elects the plurality winner, 
almost always. Top two runoff results in a "comeback election" in 
about one out of three runoffs, apparently. That's not happening with 
the RCV elections in the U.S. (the nonpartisan ones: the easy and 
relatively predictable vote transfers in partisan elections do result 
in some comebacks -- but still, often, majority failure.)

Further, the possibility of write-in votes in runoff elections -- 
which are allowed in some places in the U.S.! -- means that, given 
sufficient preference strength, Condorcet correction becomes 
possible. All the supporters of the Condorcet winner have to do is 
realize that it could happen and write that candidate in. Yes, 
tricky. That's why it takes "sufficient preference strength."

Preference strength explains a great deal about voting systems that 
we have missed when we only looked at rankings. Not finding a 
Condorcet winner doesn't mean much if that winner is one through weak 
preferences, and the alternative is strongly preferred by those who 
prefer him or her. The Condorcet supporters will presumably accept 
the other candidate.

Using advanced methods with the primary and runoff, instead of trying 
to replace the runoff system with a single ballot method, will give a 
truly advanced overall system, better than *any* single ballot method.