[EM] piling on against IRV

[Abd ul-Rahman Lomax]

At 01:42 PM 5/9/2010, robert bristow-johnson wrote:
>On May 9, 2010, at 12:29 PM, Kathy Dopp wrote:
>
>>>From: robert bristow-johnson <rbj at audioimagination.com>

[Kathy Dopp had written:]

>>>>In truth, IRV and STV are an enormous step *down* from existing
>>>>plurality voting,
>>>
>>>...
>>>
>>>>IRV/STV also finds majority winners far *less* often than does any
>>>>primary/general or top-two runoff plurality election system, unless
>>>>you apply your deceptive creative new definition of "majority" as
>>>>not
>>>>a majority of all voters who cast ballots, but of all voters whose
>>>>ballots are not eliminated from consideration by the final counting
>>>>round (some of them after *not* having had all their choices counted
>>>>whenever a subsequent choice was eliminated prior to a higher
>>>>choice.
>>>
>>>so how does delayed TTR solve that problem?
>>
>>In TTR, every voter is allowed to vote and all their votes are
>>counted.  I'm surprised you didn't know that.

And she told the truth. A great deal of propaganda has been 
circulated about top-two runoff, based on the false analogy between 
TTR and IRV; we should remember that "Instant Runoff Voting" was a 
name invented for political purpose, to sell single-winner STV. But 
Robert's Rules of Order, when it discusses "preferential voting," and 
it gives an STV counting method as an example, is very aware of the 
difference. First of all, the default under Robert's Rules is 
repeated ballot, and all previous ballots are iarrelevant to the 
current one. TTR is itself an invention, a compromise, and it could 
be a lot better; the problem is elimination.

>what voter, in the IRV election in Burlington 2009 was not allowed to
>vote.  other than 4 ballots (out of 8980) that were machine rejected
>(3 of those ballots were later examined and hand counted) were not
>counted?

They were allowed to vote but their vote was not counted (in what 
determined the winner).

When I analyzed the IRV (RCV) elections in San Francisco, using 
ballot information to see what would have happened with Bucklin, an 
IRV supporter pointed out that voters would vote differently if the 
method were different (in this case due to Later No Harm violation by 
Bucklin, allegedly -- I think it's been exaggerated). But here, when 
comparing IRV and Plurality, it will be assumed that voters will vote 
the same, i.e., for their favorite, no matter what the political 
context is. Which is preposterous. Some will. Not all.

In Burlington, there were three leading candidates, the Progressive, 
the Republican, and the Democrat. Burlington is an unusual town for 
the U.S.: the leading party is the Progressive Party, the Republicans 
are second, with the Democrats close behind them.

Republicans don't win in Burlington, because the Progressives and the 
Democrats together have a strong majority. With Plurality (and top 
two runoff), the voters know this. But with Top Two Runoff, a 
Republican can safely vote for the Republican, and then be assured of 
being able to vote for the Republican or the Progressive in the 
runoff. Top Two Runoff encourages sincere voting in the primary. (But 
it can break down, and, in fact, it is breaking down in this 
situation, since the Condorcet winner is probably the Democrat; this 
is Center Squeeze, which afflicts both IRV and TTR).

So with IRV, the Republicans, having been snookered by the promises 
that this was "like" runoff voting, and that they could safely vote 
for their favorite, did just that, and so the Democrat was eliminated 
before the Republican. Because the Republican was then facing the 
Progressive in the last round of counting, all second-choice votes 
for the Democrat were not counted.

This is important. In a real runoff election, if the voters care 
about the result, their votes in the runoff are counted, all of them. 
In an "instant" runoff, if you happen to prefer a later-eliminated 
candidate (or later-losing, as in this case), your lower choice vote 
is not counted. It's "hidden" under your first-preference vote.

So the Republicans "voted," for sure. But they voted depending on a 
belief that the method would protect them for voting sincerely, and 
allow their lower ranked votes to be counted if their first 
preference wasn't going to win. The "Later No Harm" protection of 
IRV, however, prevents their votes for the Democrat from being 
counted. Their candidate wasn't going to win, though, so this LNH 
protection was completely useless, and caused them a substantial 
losss of satisfaction. Had they reversed their preference, even only 
some of them, the Democrat would have won, an outcome that apparently 
they would have preferred.

>>TTR - all voters are allowed to participate
>
>and which voters, in the IRV election, were not allowed to participate.

Semantics. Burlington was unusual, they did have enough ranks. It 
appears that some Burlington voters in an election there that I 
studied imagined that if they ranked a candidate last, they were 
voting against this candidate. No. They were voting for the candidate 
and against all write-in candidates. Not that it mattered. Bucklin is 
much clearer: if you vote for a candidate in Bucklin at any rank, you 
are voting for the election of the candidate. If you are not willing 
to do that, you leave the candidate unranked. It makes a big 
difference if Bucklin were used as a primary method in a runoff 
system. Want to ensure that there is no runoff? Vote for all the 
candidates in your lowest rank!

>>IRV/STV - the more candidates who run, the fewer voters can
>>participate in the final counting round, given the US system of
>>allowing up to 3 ranks on a ballot.
>
>you are ignorant of the fact that in Burlington, all 5 candidates were
>ranked (at least they were on the ballot i had).  Burlington is not
>San Francisco.  we evidently have stricter ballot access laws.  i
>don't know what Burlington would have done if there were 25 mayoral
>candidates on the ballot.

There are not many candidates on the ballot for the Mayor of San 
Francisco. They get over twenty candidates for certain supervisorial 
districts. In general, advanced voting systems do encourage more 
candidates to file, but the damage is already done in places that use 
top-two runoff, which is generally cnsidered to allow sincere voting 
in the primary. It isn't completely safe, though.

>in comparison, i have seen 3 different TTR elections for City Council
>in Burlington.  none had more than 55% turnout on runoff day (in
>comparison to the number of voters that came on the first election
>day).

>   the IRV election had 93% of the voters participating in the
>final round.   93% turnout is a lot better turnout than 55%.

This is confusing turnout with voting. IRV supporters often stir up 
this confusion. Voters turn out to vote when they care about the 
outcome. A higher turnout does not necessarily indicate better 
results, better overall voter satisfaction; in fact, it can mean the 
opposite, when top two runoff misses the Condorcet winner, as with 
some famous elections: Lizard vs Wizard, where a former Grand Wizard 
of the Ku Klux Klan edged out the candidate who was probably the 
Condorcet winner, bringing a runoff between the Wizard and the 
frontrunner in primary votes, a candidate called "The Lizard," 
because of some serious scandals. The prospect of the Wizard winning 
was enough to bring out voters in droves; the runoff turnout was 
higher than the primary. The Wizard got a few more votes in the 
runoff than in the primary, but the rest of the voters held their 
noses and elected the Lizard.

What normally happens with top two runoff, though, is that the best 
two candidates go to the runoff. There will be lots of voters who 
have no strong preference between these candidates. So they don't 
turn out to vote. These voters will not be offended with either 
outcome. Those who care, vote. And that has always been true with 
democratic process, in spite of a lot of propaganda about how it is 
always better to vote than to not vote.

So... suppose this Burlington election had been Plurality. The 
Republican voters know that the Republican never wins. So they can 
choose to vote for the Democrat, whom they prefer to the Progressive, 
say, which might cause the Democrat to place second. Then they can 
vote in the runoff, and the Democrat will win. If the method were 
Bucklin, they'd really have some clear and easy ways to vote that 
would lead to improved outcome. Or they vote for the Republican in 
the primary. Some will do one of these, and others will do the other. 
If the Republican comes in second place, which is what will happen if 
all Republicans vote sincerely, the Republican and the Progressive 
will face each other.

Who will turn out to vote in the runoff? Hard to say. Democrats may 
be split and the Democratic vote, besides being low turnout, may make 
no difference.

Now, Mr. Bristow-Johnson is comparing a mayoral election with city 
council elections. He assumes, apparently, that turnout would be as 
high if it were IRV for City Council, which it wasn't. Experience 
with top two runoff shows that certain elections have high turnout in 
a runoff. In other jurisdictions, the primary is held before the 
general election in November, and there is a runoff in the general 
election if needed. Turnout in the runoffs tends to be about equal to 
that of the primary, in spite of what many might expect.

Top two runoff is a long-time major election reform that FairVote has 
been busy dismantling. Instead, real election reformers would have 
focused on using a better primary method, that is more likely to find 
true majorities (IRV only does it rarely when runoff rounds are 
needed), that will include the true top two (or three, under some 
conditions) candidates. And some very simple methods exist; the one 
that would work easily and well in Burlington is Bucklin. It would 
allow the Republican to vote in first rank for the Republican. Some 
would add a lower ranked vote for the Democrat, and it would be 
enough to either cause the Democrat to win, or the Democrat would 
make it into the runoff.

>every
>voter that expressed an opinion of at least one of the two candidates
>that made it to that final round participated in the actual choice of
>the elected candidate.

Notice how contorted this phrasing is. This is the bottom line: IRV 
does not count many of the votes cast. They are discarded, they are 
made moot before being counted. In the case of Burlington, because 
the ballot images are available, they have been counted, which is how 
we know what happened. The method ignores them. IRV decided that the 
Democrat could not win before a large chunk of votes for the Democrat 
were made visible.

It may be said, indeed, that there is an analogy with top two runoff, 
but in top two runoff, no matter what the first rank vote was, the 
voters -- a new set of voters! -- would be able, in the runoff, to 
make their choice. I support allowing write-in votes in runoff 
elections, and using an advanced method, spoiler-free, which would 
allow a write-in campaign to make an effort to improve the outcome 
without spoiling the election; if an advanced method is used in the 
primary, it will not only avoid a lot of runoffs, it will gauge the 
possibility of success of a write-in campaign. Dark horses become 
visible with a good method.

>>It's a very simple concept to understand.
>
>which you evidently don't.  your arrogance, Kathy, is greater than
>your ignorance.

The insult reflects on the writer, not on his target.

>   you greatly underestimate the people you talk to here
>on this list.  no one here is using those canards as criticism of
>IRV.  we know what they are.




>we all expect that people fill out their ballots to express their
>political interest (or, perhaps, they are trying to vote
>strategically).  none of us are stupid enough to buy into the canards
>that you and the IRV opponents repeated over and over again that
>somehow IRV "disenfranchises voters" or doesn't count their vote.

"None of us"? Robert, speak for yourself. What Kathy wrote was 
basically correct.
[...]

>>Where is your data or evidence to support your claim that the voters
>>in Burlington, VT would vote for their 1st choice IRV/STV candidate if
>>the contest were plurality?
>
>what is your evidence that they would not?  the burden of proof lies
>on you for that.  *you* need to show that, for a significant number of
>voters, their single vote on the "traditional ballot" would be
>different than their favorite candidate.

When people vote sincerely in Plurality, and when there are more than 
two candidates who are at all viable, the results frequently lead to 
majority failure, but in a Plurality system, the candidate with the 
most votes wins. This is well-known to lead to a two-party system, 
and the bulk of voters, even if they might prefer a different 
candidate, will vote for one of the top two. It's circular, for sure, 
but that's how it works, and that Bristow-Johson is ridiculing what 
is common knowledge, again, simply reflects on him.

Plurality usually works, and the way that it works is by defering the 
"primary" to party political process, where parties are involved, as 
they are in Burlington. So the public election is really a runoff 
between the Democrat and the Republican, in most places. This, of 
course, breaks down when the distance between the Republican and the 
Democrat becomes too far, and someone appears in the middle, but not 
yet with enough popularity to win from first-preference, sincere 
votes, or perhaps not considered to have a chance.

There are simple fixes, but by pushing IRV, FairVote has possibly 
damaged the chances for true election reform in Burlington for many years.

I've been meaning to respond to Mr. Bouricius' comments, which are, 
in certain respects, fair enough, but which do neglect the history of 
anomosity between FairVote and the election methods community.



>>  It seems to me that many voters would
>>have recognized that they needed to vote for one of the likely top two
>>winners in a plurality contest, whereas in IRV/STV they were fooled by
>>rhetoric such as Bouricius' into thinking they could vote honestly
>>without hurting their 2nd choice candidate.
>>>yeah, but the Dems have less in common with the Repubs than they have
>>>with the Progs.  that's what the numbers say.  that's why, in the
>>>final IRV round the vast majority of Montroll votes got transferred
>>>to
>>>Kiss than those that were transferred to Wright.
>>>
>>>>who would have both gotten a far better result
>>>
>>>NO THEY WOULDN'T!  That is Your Lie.  the interests of the Democrats
>>
>>Oh.. I see that you are certain that you read the minds of all
>>Burlington, VT voters better than I do, so that you can make the
>>unlikely and unsupported claims that:
>
>__________________________________
>
>all the below is crap.

Most of what was below was written by Bristow-Johnson. He should find 
somewhere else to crap. Who is this jerk?