[EM] The deception at "irvfactcheck" Part 3.

[Abd ul-Rahman Lomax]

http://irvfactcheck.blogspot.com/p/there-is-lot-of-miss-information.html


>
><http://irvfactcheck.blogspot.com/p/there-is-lot-of-miss-information.html>Myths 
>and Facts about IRV
>
>
>

Continuing from Part 2

>1.3 Do voters like using IRV?
>
>Yes. Voters overwhelmingly prefer IRV to the previous election 
>method, according to exit polls conducted in U.S. cities that use IRV.

Conducted by whom? Check it out. Typically conducted by FairVote or 
FairVote enthusiasts.

>  In every case, a substantial majority of voters preferred IRV 
> compared to those who preferred the old system or had no opinion. 
> Excluding those who said the methods were equal or who had no 
> opinion, the gap between those favoring IRV and those favoring the 
> old system ranged from 71% vs. 29% in the lowest city to 89% vs. 11%.
><http://www.fairvote.org/?page=2170>[More information and links to 
>the various exit polls]

I'm not going to go into much detail here. What really has mattered 
is that IRV has been voted out after being used, in a significant 
number of places. That the method was difficult to use or confusing 
was an argument raised. But it seems that other factors may have been 
as important or more important. The cost of tabulating the vote. The 
quirkiness of the results. The uselessness compared to plurality 
(same results!). The failure to obtain real majorities (which was 
important in most of the jurisdictions). If these were exit polls, 
the voters did not know the results yet. They were new to IRV and may 
not have realized what the implications were of how they had voted.

I've seen, in the actual votes from Burlington, signs that voters 
really did not understand how to vote with IRV.

I forget the actual numbers, but suppose there were five candidates 
on the ballot, plus a space for a write-in. The ballot allowed you to 
rank five candidates. So some voters ranked all five candidates. That 
meant that they were voting for *every candidate* against any 
possible write-in vote. Really? Even the absolute worst of them? I 
think they simply didn't understand.

>The City of Cary, NC used IRV in a pilot program authorized by the 
>state. Both election officials and voters were extremely pleased with IRV.

Not what I've read from them!

>Cary conducted a general public opinion survey of residents, and 
>included some questions about IRV. Residents favored IRV over the 
>former method by huge margins. The survey results can be viewed by 
>clicking 
><http://www.sitesubplans.townofcary.org/__shared/assets/irv17690.htm>here: 
>(note that this survey was of residents, not just voters, and so 
>included many people who had no exposure to voting place voter education.)

Those poll results are being presented wtih rose-colored glasses by 
FairVote. And the respondents in that poll were told some 
misinformation about how much money would be saved by not holding a 
runoff. First of all, runoffs are not always held. Secondly, 
neglected in that was the extra cost of tabulating IRV when it's needed.

22% of respondents said that they "did not understand [IRV] at all."!

I was unable to verify who ran the poll. I found the following on the 
Cary web site, presented to the City Council in February, 2009:

>Dr. Michael Cobb, Professor at NC State University, designed an exit 
>poll to evaluate how voters in Cary felt about IRV.  The exit poll, 
>which was managed by Bob Hall, Director of Democracy North Carolina, 
>contained interviews with more than 1600 voters from Cary Town 
>Council Districts B and D.  The Board of Elections approved the 
>conducting of the poll; however, the Board did not make any findings 
>regarding the poll results.  Dr. Cobb analyzed the results of the 
>polls.  The results were sent to the State Board of Elections. Key 
>findings of the survey were:
>
>Of those with a preference, 72 percent of Cary voters said they 
>preferred IRV while just 28 percent said that they preferred voting 
>for a single candidate.
>
>Almost everyone (96%) reported it was at least "somewhat easy to 
>understand" the IRV ballot with 82% agreeing that it was "very easy" 
>to understand.
>
>Most voters (69%) actually utilized the option of ranking at least 
>two of the candidates for city council
>
>Among the minority of voters who did not rank more than one 
>candidate and gave a reason why (480), only 29% (139) said the 
>reason for not indicating a second choice was that they were 
>confused about how the rankings would be used.
>
>Voters were more likely to rank candidates in District B, which was 
>the most competitive race where all 3 candidates failed to win an 
>outright majority; voters in District B were more likely to prefer IRV.
>
>The study found no significant differences between different types 
>of voters in their understanding or preference for IRV:  whites and 
>non-whites, males and females, lower- and higher-income voters all 
>evaluated IRV roughly equally.
>
>Outreach efforts to inform voters ahead of time about IRV were 
>largely successful.  Seventy-six percent said they knew they would 
>be asked to rank their preferences before coming to vote that 
>day.   And those who reported knowing about IRV in advance were more 
>likely to rank more than one candidate and to prefer ranking 
>candidates over voting for only one candidate.

I believe Bob Hall is a FairVote activist. 31% of voters did not rank 
an additional candidate? I've seen lower bullet-voting rates than 
than in Bucklin election records.

http://www.townofcary.org/Sunshine/Agendas___Minutes/Town_Council/Action_Agenda_Archive__Abbreviated_Minutes_/2009_Action_Agenda/aa031209.htm

E.   PUBLIC HEARINGS

1.   Public hearing on (a) the method of election for the Town of 
Cary beginning with the 2009 municipal election and (b) whether to 
utilize Instant Runoff voting in the 2009 municipal election. 
(Approved pursuing the non-partisan plurality method of election; the 
legal process will begin at the 3/26 council meeting. The council did 
not make a decision on Instant Runoff Voting. In order to bring up 
the issue of Instant Runoff Voting on a future agenda, two council 
members must submit a request in writing to the town clerk.)

There appears to be no move to use IRV again in Cary. But Joyce McCoy 
would know more, I'm sure.

Back to the blarney:

>1.4. But hasn't IRV been repealed in numerous jurisdictions?
>
>Not exactly. In the 21st century, voters in X jurisdictions have 
>voted for charter amendments or statutory changes to move to IRV. 
>Voters in two jurisdictions have passed charter amendment to repeal 
>it: Pierce County (WA) and Burlington (VT). [Explanations]

Notice: they haven't figured out the explanations yet. Notice as 
well, they limit this to "the 21st century." IRV was used quite a few 
times in the U.S. in the 20th century. What happened? FairVote hardly 
ever mentions this. Are they going to tell us why the voters repealed it?

>Opponents like to list other repeals, but that is inaccurate [Aspen and Cary]

Sure. It wasn't repealed in Cary because it was a pilot program. They 
had the possibility of continuing it in 2009, and didn't. Why not? 
Aspen is more complicated.

>They even like to keep saying Georgetown University has repealed IRV 
>because they happened to come across a news story saying it did. But 
>a few weeks later, Georgetown students restored IRV nad have used it 
>ever since, along with more than 55 additional colleges and 
>universities? Will all those schools keep IRV? Probably not. But we 
>expet the overwhelming number will keep it and support it. That's 
>the same with IRV in city elections.

Not if history is any guide. IRV is an expensive equivalent, in most 
current applications in the U.S., to Plurality. This isn't a "Fact 
Check." This is pure FairVote propaganda. I really don't care what 
happened at Georgetown. Notice the contradiction: they imply that 
people are mistaken about Georgetown repealing IRV. But apparently it 
was indeed repealed. But later restored.

There are an enormous number of elections every year in colleges and 
universities. Only a very small number use IRV, and this is likely 
because of intensive efforts by FairVote to promote the method there. 
I came to know one of these student activists well. He had basically 
been brainwashed, coached in how to present IRV to make it seem 
positive. He was pretty smart, though. It didn't take much to get 
through that conditioning.

FairVote has made a big deal about the use of IRV in Takoma Park. 
Takoma Park is a small bedroom community in Maryland, it's where the 
Executive Director of FairVote, Rob Richie, lives. They need IRV like 
they need a fish bicycle, most elections are uncontested or have only 
two candidates. IRV has made no difference there at all, except to 
FairVote, because it was one of their first "successes."

My guess is that they don't really care what happens in the long run, 
just as long as the bubble lasts long enough to get Rob Richie 
through retirement....

Most people with an undertanding of the situation seem to think that 
IRV is doing serious damage to the cause of election reform in the 
U.S. When a jurisdiction tries IRV and discovers that it really 
doesn't help, or, as in Burlington, it does real damage, they are 
soured on the concept of election reform for a long time. Maybe. 
Let's hope that this does not prevent the consideration of better 
methods, but FairVote is doing everything it can to deny that better 
methods exist. There are objective ways to judge voting system 
performance. You will not see a trace of this from FairVote.

>IMPACT OF IRV
>2.1 Does IRV address the "spoiler" problem as well as separate 
>runoff elections do?
>
>Yes, but better. IRV solves the spoiler problem better than separate 
>runoffs and at least as well as any alternative voting method. A 
>"spoiler" is a negative term for a minor candidate with not chance 
>of winning the election. Under plurality rules such "spoiler" 
>candidacies can throw an election to a candidate the majority of 
>voters oppose. IRV tackles the "spoiler problem" in a manner similar 
>to a traditional two-election runoff, except that, due to its 
>sequential elimination procedure, it is more effective at 
>eliminating the spoiler effect.
>
>Under a traditional two-election runoff, if there are several 
>candidates in the first round who have similar policy views, it is 
>possible for them to split the majority in such a way that none of 
>these similar candidates reaches the runoff round, while two 
>candidates with substantial core support but very narrow appeal 
>advance. Under IRV, voters who divided among these similar 
>candidates will see their votes automatically re-unite for the 
>strongest candidate in the group, and thus advance to the final 
>runoff. Some advocates for other alternative voting methods have 
>claimed that IRV does not solve the "spoiler problem" 100%. IRV 
>solves the problem as well or better than the alternatives.
><http://www.fairvote.org/?page=2491>[Examples of spoiler scenarios 
>under Plurality, Range, Approval and other voting methods]

Notice that to "spoil" an election with PLurality, the winner must 
fail to gain a majority. Now, IRV elects without majorities, the way 
the method has been implemented (as a runoff substitute). Voters, we 
should remember, are free to vote how they choose, and it's rather 
obvious that if voters vote the way they vote with Plurality, with 
IRV, a "spoiler" can result in the election of a candidate who would 
be beaten by another in a direct face-off. But IRV frequently does 
this, and in more than one way. IRV, as to its positive function, is 
practically identical to top-two runoff, but without the critical 
re-examination by the voters. There are far better ways to improve 
top-two runoff than by collapsing it to a single ballot no matter 
what, which is the IRV approach. Bucklin was also designed originally 
to make runoffs unnecessary, but it, too, sometimes failed to find a 
majority. From my analysis, it appears that it's more likely to find 
majorities, real majorities, than IRV.

If a majority is considered important (and IRV legislation often is 
prefaced with a preamble about how important finding majorities is), 
then ... a runoff is sometimes a necessary cost. Bucklin, because it 
counts all the votes, is more efficient at avoiding runoffs than IRV 
used as a primary method would be, and probably, because of the 
ability to express a Favorite vote, in addition to other approvals, 
it would be more efficient than Approval. In a sense, using a 
three-rank Bucklin system for primary and runoff would be like 
setting uo to hold as many as six runoff elections, using Approval 
voting, three at a time, in which the voters can set which round to 
introduce their additional approvals, thus showing their preference 
strengths as well as preferences. Bucklin can be extended to use more 
than three ranks and becomes even more accurate and flexible. It can 
use a full Range ballot, which is the most expressive possible 
ballot, allowing voters to, if they wish, accurately express how they 
rate candidates.

(A Range ballot could actually be used to run IRV; just treat it as a 
ranked ballot. But that is tossing away possibly valuable 
information. Ratings ballots, where candidates are independently 
rated, can serve as ranked ballots, but preference strength 
information is lost.

FairVote is here mentioning that critics claim that IRV doesn't solve 
the spoiler problem "100%." That's true. Why don't they explain what 
these critics mean. You would not have a clue, and you would think 
from this that it's all relative, just a question of degree, and that 
IRV does "as well or better." That's a lie. IRV handles the "first 
order spoiler effect," I've called it. This is the minor, no-hope 
candidate problem. There are other methods which also solve this 
problem, it's actually easy to solve, IRV does it well. But then IRV 
sets up a much worse problem, "center squeeze," where, by leading in 
first preference votes, a candidate who will go on to lose causes a 
much more popular, overall, candidate to win. Popular in the sense of 
being preferred to every other candidate by as much as two to one.

Now, as to 'examples of spoiler scenarios,' the reference is to a 
FairVote document, which is at least better referenced than most 
FairVote work, but it's still blarney. Consider this:

>IRV is fully resistant to classic spoiler scenarios in which a minor 
>candidate splits off votes from an otherwise winning candidate. 
>However, IRV can still suffer a dynamic similar to the spoiler 
>dynamic when there are three or more candidates with strong support. 
>In such cases it may sometimes be possible that a voter's support 
>for a favorite candidate rather than a "lesser-of-two-evils" 
>candidate may result in the favorite making it to the final runoff 
>instead of the compromise candidate, and yet the favorite candidate 
>may be unable to win in that runoff against a less-preferred choice, 
>whereas the compromise candidate could have won. There are no known 
>examples of this scenario playing out in real world IRV elections, 
>but it is at least conceivable.

The first sentence isn't correct. It is not "fully resistant." But 
I'd agree that it's strongly resistant to the spoiler problem.

Here, notice, their own paper is acknowledging a problem with center 
squeeze. "No known examples" was preposterous in substance. IRV 
behaves very similarly to top-two runoff, a point which they like to 
make when it makes the method seem familiar to people, but top-two 
runoff suffers from this same center squeeze problem. So they would 
know that it's possible, and even somewhat common, there are quite a 
number of famous elections where it happened. And then, with IRV, 
Burlington. It actually happened, and, folks, this is why Burlington 
dumpted IRV, almost certainly. Center squeeze can happen in just the 
situation that they described, where there are "three or more 
candidates with strong support." That was Burlingon, and it is a 
routine situation there, it could certainly have been predicted that 
it was possible.

The Fairvote document doesn't cover Bucklin at all, and in order to 
impeach the spoiler behavior of Approval and Range, it makes up 
internally contradictory situations. I.e., they assume that 
supporters of a spoiler will bullet vote for their favorite. But they 
can do this with IRV quite as well... Bucklin completely resolves 
this by allowing voters to show their first preference without 
ambiguity, which is the problem with Approval. Range likewise allows 
the showing of the first preference with only a tiny loss of voting 
power in the "main show." But the real fix for the center squeeze 
effect is to use an advanced method, such as Bucklin (or Range!) in a 
primary election in a runoff system. If voters err on the side of 
bullet voting, it causes majority failure and they get another 
chance. With proper method design, the compromise winner will 
definitely make it into the runoff and will win (assuming that the 
support is deep and not just some artifact of how people happened to 
make "strategic voting decisions.) There is no doubt that in such a 
system, the Democrat would have won in Burlingon, as the IRV votes, 
when analyzed, showed was the majority choice.

FairVote has been denying what voting systems experts have known for 
more than a century: IRV is a lousy single-winner method. There are 
much better methods, from the simple repeated ballot used in 
deliberative bodies, which can be approximated for public use by 
Bucklin, which seems to have been proposed by the Marquis de 
Condorcet before 1800, but was reinvented by James Bucklin in the 
U.S. and first used in 1909, and which swept the country as 
practically a fad, heavily supported by political scientists at the 
time -- and then was swept away in a political backlash that wasn't 
about how the method actually worked but seems to have been more 
about the fact that it did work: it elected, in the first election, a 
Socialist as Mayor of Grand Junction, Colorado, and Bucklin was 
definitely not welcomed, either, by the "Democratic" party machine, 
for the last things the party bosses want are voting methods that 
allow voters to make the choices instead of them.

Bucklin voting is not, technically, "Condorcet compliant," but that 
is a huge question. As Warren Smith wrote at one point about Range, 
it may be "more Condorcet than Condorcet," which was probably 
referring to an ability to maximize overall satisfaction as distinct 
from the raw vote counts. Bucklin only fails to elect the "Condorcet 
winner" if it fails to find a majority, sometimes, or it finds a 
majority for more than one candidate. Those situations can be fixed 
if Bucklin is used in a runoff voting system, such that if it does 
still fail to find the Condorcet winner, it is because there is no 
practical difference between that candidate and another. It would 
have to be very close, and a very unusual situation.

The huge center squeeze problem of IRV isn't about a "close 
election." It is about IRV ignoring many or even most of the votes 
cast. Whether your vote cast for the compromise winner (preferred by 
as much as two-to-one over all other candiates) is ever counted 
depends on whom you prefer at a higher rank. IRV does *not* treat all 
votes equally. It does treat voters equally but only in the sense 
that all ballots are counted according to the same rules.