[EM] Does IRV elect "majority winners?"

Abd ul-Rahman Lomax abd at lomaxdesign.com
Sun Jan 4 13:10:16 PST 2009


At 05:26 PM 1/3/2009, Kathy Dopp wrote:
> > Date: Sat, 3 Jan 2009 14:29:23 -0500
> > From: "Terry Bouricius"
> > However the norm in governmental elections is to discount all abstainers
> > from the basis, regardless of the manner of their abstention.
>[...]

>It seems that IRV/STV proponents have no limit for the convolutions
>they'll use to justify their own unique definitions of commonly
>understood concepts like "majority" in order to mislead the public
>into supporting IRV/STV.

It should be noted that Terry is not merely some loose cannon; we see 
preposterous arguments from such, from time to time, here on the 
Election Methods list and elsewhere.

He is a co-author with Rob Richie, the Executive Director of 
FairVote. He has been listed as a consultant to FairVote.

Here is what he says about himself at http://www.fairvote.org/?page=2224

>I served ten years on the Burlington, VT City Council, including a 
>term as president. I then served ten years as a member of the 
>Vermont House of Representatives, where I drafted legislation on 
>instant runoff voting. After leaving the legislature I served on the 
>state board of the League of Women Voters of Vermont. 
>Professionally, I am an elections administration consultant, having 
>administered elections for non-profit organizations and consulting 
>on ranked-voting election methods. I am a senior policy analyst for 
>FairVote: The Center for Voting and Democracy, and have given 
>testimony and consulted on IRV with Republicans and Democrats in 
>over a dozen states and numerous municipalities. I designed the 
>ballot, the vote tabulation procedures, and voter education plan for 
>Burlington., Vermont for its 2006 IRV election. [...]

>In 1998 the Vermont House of Representatives passed a resolution 
>creating a Citizens Commission to Study Preference Voting. This 
>Commission had members from across the political spectrum, including 
>former legislators, and others. The Commissions final report 
>(available at http://fairvote.org/irv/vermont/index.html ) 
>unanimously recommended the adoption of instant runoff voting for 
>all statewide elections. The key finding of this commission was that 
>plurality election rules in common use through out the U.S. cannot 
>accommodate more than two candidates without risk of "spoiler" 
>situations resulting in the true will of the majority being 
>thwarted. IRV was deemed the best solution.

The final report shows no signs of participation by people with a 
broad knowledge of voting systems. If we look at the Executive 
Summary, we find:

>This Commission recommends the adoption of Instant Runoff Voting 
>(IRV) for statewide elections as a remedy to a potentially serious 
>defect in Vermont's election laws. Vermont's plurality election 
>rules allow for the election of a candidate with the most, but less 
>than half, of the votes, even if the majority of voters oppose this 
>candidate and prefer a different one. This is a fundamental defect 
>that violates the most basic precept of democracy: majority rule. 
>Today the plurality problem in the case of the offices of Governor, 
>Lieutenant Governor and Treasurer is resolved by falling back on the 
>legislature. If no candidate receives a popular majority, a secret 
>ballot election by the members of the General Assembly, rather than 
>the voters, decides the race. The General Assembly has had to choose 
>state officers 69 times, often electing a candidate that had come in 
>second in popular votes, in one case electing a third-place 
>candidate who had received 3% of the vote, and in another failing to 
>elect a Governor altogether. In 35% of all election years, at least 
>one statewide race has had a result with no majority winner.

Sure. Problem is, IRV doesn't fix this with any reliability, and it 
can *still* elect a candidate when "the majority of voters oppose 
this candidate" -- and their votes express this!

>For most of Vermont's history, a majority vote was required to win 
>all single-seat elections, and runoff-like re-votes were common. It 
>wasn't until 1940 that re-votes were completely done away with, and 
>a plurality of first-round votes was deemed sufficient for election 
>to any office other than Governor, Lieutenant Governor or Treasurer. 
>The inconvenience of re-voting was felt to be a bigger problem than 
>the risk of undemocratic outcomes. Since IRV eliminates the 
>inconvenience of re-voting, there is no longer any reason to use an 
>election process that allows for the defeat of the candidate 
>actually preferred by a majority of voters.

I'd interpret the matter differently, of course. What is happening 
here is that democratic requirements are being discarded for 
convenience. But, once again, the concept of "majority" is being done 
violence to. This comment is in the context of a recommendation for 
Instant Runoff Voting, *as if* it is not such an election method as 
described above. There are better single-ballot voting systems than 
IRV, but, of course, *no method* can guarantee a majority unless it 
coerces voters or prevents them from expressing themselves freely.

A description of the method was provided:

>This process of dropping off bottom vote-getters and transferring 
>their votes to their supporters' alternate choices continues, until 
>a candidate gets a majority, or only one candidate remains.

"Majority" wasn't defined. That there would be exhausted ballots 
wasn't mentioned. This is far from a neutral report, this report was 
a piece of crafted propaganda, inconvenient facts that *will* be 
mentioned in a true neutral report are just glossed over.

>IRV is preferable to a two-round runoff election in that it saves 
>money, assures that the deciding election will have maximum voter 
>turnout, and does not face constitutional problems. In a single 
>election, IRV ensures that a candidate actually preferred by a 
>majority of voters can win, and eliminates the existing problem of 
>multiple candidates splintering the vote.

Note the incorporated assumptions.

(1) IRV will save money.
(2) IRV elections will maximize turnout.
(3) IRV does not face constitutional problems.
(4) IRV ensures that a candidate actually preferred by a majority of 
voters can win.
(5) IRV eliminates the existing problem of multiple candidates 
splintering the vote.

Some of these are probably false. Some are arguably true, but other 
methods would do a better job.

>IRV tends to reduce negative campaigning. One reason for this is 
>concern over alienating voters who would not give a nasty campaigner 
>a second-preference vote, which that candidate might need to win. 
>Although negative campaigning has not yet become a widespread 
>problem in Vermont, IRV may help protect campaign civility here.

Remember, this is a Commission report. Was there evidence for this 
"negative campaigning" claim? I haven't read the whole report yet, I 
will. But this claim has been made, over and over, it is a *sales 
point*, which the Commission simply repeats as if it were a known 
fact. It probably isn't true, at all.

Consider the situation that is normal: two major candidates, and some 
minor ones. The minor candidates *aren't* going to win the election, 
normally that is a given -- that there are exceptions, rare ones, 
doesn't change the substance of this. Are the top two going to 
attempt to get each other's supporters to vote them second 
preference? Those are useless votes, you know, they won't ever be 
counted even if they are cast, under the IRV rules. No. There is no 
incentive for them to abstain from negative campaigning, except the 
same incentive that already exists: done clumsily, it can make you 
look bad, and you will lose undecided voters. From a theoretical 
standpoint, there is no reason to expect IRV to have a major impact 
on negative campaigning, and, from the San Francisco experience, it 
appears that it doesn't.

>IRV will not increase the cost of holding elections, other than the 
>minimal cost of conducting recounts when there is the lack of a 
>first-choice majority. There would be some small transition cost for 
>a voter education campaign.

I hope they are not still claiming this! IRV has had enormous 
implementation costs and serious counting costs. It *might* pay back 
this cost, over some long time. But there is another method of 
preferential voting, variously called American Preferential Voting or 
the Grand Junction System or Bucklin, which is much easier to 
canvass, which used the same ballot, essentially, as the three-rank 
RCV ballot that has actually been used, and which probably resulted 
in about the same degree of additional preference votes as IRV does, 
and which is *more* likely, from such votes, to find majorities in 
nonpartisan elections; it's probably about the same with partisan 
ones, where voting follows relatively predictable patterns. So why 
isn't this system being considered?

The last sentence of this report:

>IRV assures that a candidate preferred by the majority of voters 
>will not be defeated by a candidate preferred by a minority, and 
>strengthens Vermont's democracy for the next century.

It's a lie. IRV makes no such assurance; the method, as described and 
as was attempted in Vermont, doesn't seek majorities except in the 
first round of counting. It could easily be modified to do so, and 
this is, in fact, how Robert's Rules of Order describes it. However, 
this, then, would mean that it wouldn't eliminate all runoffs, and 
the cost savings are a major part of the campaign arguments.

Now, back to Bouricius' testimony before the Colorado House:

>The first use of IRV in Burlington was in 2006.
>
>By all accounts, the Burlington IRV election was extremely 
>successful. Exit polls conducted by a political science professor at 
>the University of Vermont found that voters overwhelmingly preferred 
>IRV to the former voting method. Some people had worried that voters 
>might not be able to handle using ranked ballots leading to lower 
>turnout or many spoiled ballots. However, turnout was higher than 
>recent mayoral elections, and among those voting in the mayoral race 
>99.9% of ballots were valid.

No, 99.9% of ballots weren't "valid," not if, as Bouricius has been 
arguing, the voters with exhausted ballots abstained. Did they cast 
valid ballots or not? If not, then it's legitimate to exclude them 
from the basis for majority. But if the votes were valid, how can 
they be excluded? What actually happened in Burlington? If this 
election was, "by all accounts," "extremely successful," then surely 
it found a majority. It did not.

http://www.burlingtonvotes.org/20060307/2006%20Burlington%20Mayor%20Round1.htm
March 7, 2006:
9,788 valid ballots.
Electing 1 candidate.
Winning threshold is 4895 votes.
There were 77 invalid ballots.
(9,865 total ballots processed.)

ROUND 1 -- Tally of 1st place votes.

CANDIDATE THIS ROUND TOTAL STATUS
Bob Kiss +3809  3809   CONTINUING
Hinda Miller +3106  3106   CONTINUING
Kevin J. Curley +2609  2609   DEFEATED -- 1st round
Louie The Cowman Beaud. +119  119   DEFEATED -- 1st round
Write-ins +78  78   DEFEATED -- 1st round
Loyal Ploof +57  57   DEFEATED -- 1st round
EXHAUSTED PILE: +10  10
TOTALS: +9788  9788
Kevin J. Curley is declared DEFEATED, as his/her defeat was 
mathematically inevitable.
Louie The Cowman Beaudin is declared DEFEATED, as his/her defeat was 
mathematically inevitable.
Write-ins is declared DEFEATED, as his/her defeat was mathematically 
inevitable.
Loyal Ploof is declared DEFEATED, as his/her defeat was 
mathematically inevitable.

ROUND 2 -- Transferring all votes simultaneously from all defeated candidates.

CANDIDATE THIS ROUND TOTAL STATUS
Bob Kiss +952  4761   ELECTED -- 2nd round
Hinda Miller +880  3986   CONTINUING
Kevin J. Curley -2609  0   DEFEATED -- 1st round
Louie The Cowman Beaud. -119  0   DEFEATED -- 1st round
Write-ins -78  0   DEFEATED -- 1st round
Loyal Ploof -57  0   DEFEATED -- 1st round
EXHAUSTED PILE: +1031  1041
TOTALS: 0  9788
Bob Kiss has been ELECTED because s/he has a majority of the remaining votes.

Notice that the first page gives, correctly, the Droop quota as 4895 
votes. The Droop quota is, for a single-winner election, a majority. 
However, note, then, that with only 4761 votes, Kiss is declared 
elected, because he has a "majority of the remaining votes." I.e., a 
plurality of the votes.

What's the gap? Well, the exhausted pile is 1031 votes. These are 
voters who did not rank either Kiss or Miller. Kiss got 952 of the 
vote transfers, Miller got 880, and 1031 ballots voted for a "minor 
candidate" -- Curley was a Republican, if I've got it right, which in 
Burlington makes him a significant third-party candidate! -- and did 
not add any additional preferences. The gap between Kiss and Miller 
is 775 votes.

It is highly unlikely that any voting system (of the ones reasonably 
proposed) would have come up with a different result. If this 
election had gone to a runoff, though, we *cannot predict the 
outcome.* Miller would almost certainly be the other candidate, 
against Kiss. Kiss was (is) very popular in Burlington. But Miller 
might pull off a comeback. It *happens* with real runoff elections: 
what may have been a minor candidate, with the additional attention, 
manages to win. *This doesn't happen with IRV, unless an election is 
very, very close.*

The only problem with this election is that it shows a victory 
without a majority of votes being shown. In reality, a majority may 
have cast a vote for Kiss, *but the results don't show it,* because 
the IRV method conceals what may be a rather large number of second 
preference votes for Kiss among supporters of Miller (and the 
reverse, as well). This was a municipal election, and even when it's 
partisan, as this was, municipal elections tend to be much more about 
personalities than parties. Thus, we really should look at the ballot 
images to see what actually happened. Bucklin would count all the 
votes, that's one reason to prefer it. Same ballot; in this case, 
same result; but in this case, Plurality would have given the exact 
same result as well.

And that turns out to be typical for IRV, in nonpartisan elections -- 
and in many partisan ones as well -- it is *functionally equivalent* 
to Plurality.

Look at the results: Most of the transferred votes must have come 
from the Republican. More than half of the voters for the Republican, 
very roughly, also voted for a Democrat (Miller) and a Progressive 
(Kiss). If this doesn't clue us in that this election isn't at all 
typical for the U.S., that ought to!

The vote transfers show roughly the same preference for Kiss > Miller 
as does the overall electorate in first preference votes. That, 
again, seems to be typical for nonpartisan elections, this happens in 
Australia with Optional Preferential Voting. Where this phenomenon 
holds, IRV tends to produce the same results as Plurality; the vote 
transfers simply increase the lead, ordinarily, or sometimes reduce 
it a little, they don't reverse it.

Absolutely, from the votes, it appears that Kiss was the legitimate 
winner. That's not the problem. The problem is that no majority was 
found! It is possible, theoretically, that the voters with exhausted 
ballots could have turned it around in a real runoff. But unlikely. 
My guess at this point is that Bucklin analysis of the ballots, which 
could be done, since the ballot images are available, would show the 
Kiss *did* get a majority of the votes. But, despite all the fine 
words about "majority rule" and "finding a majority," IRV, as a 
method and as implemented, *doesn't care about "majority"* not as 
reasonably defined. (The Droop quota as mentioned on the web site.)

(Note that http://www.burlingtonvotes.org/20060307/ links to ballot 
images and format files.)

Has anyone looked to see what the overall truncation rates were? In 
any Bucklin election result, you can tell immediately, since all the 
votes are counted. In a two-party situation, we can expect high 
truncation rates among those who support the frontrunners, whether 
the method is IRV or another, such as Bucklin. But this is really a 
three-party situation in Burlington. And it's the kind of situation 
where Center Squeeze can bite. Suppose instead of the Republican 
being in third place, it had been the Democrat, and suppose that 
party affiliation were a larger factor than the Democrat's first 
place support that she came in third place, being the candidate 
squeezed in the middle, but she also might be the pairwise winner, 
i.e., she'd win a contest between herself and either the Progressive 
or the Republican. Thus this election is not far from one where a 
very poor result, *contrary to majority rule* -- as shown by the 
votes! -- could have occurred. IRV works more reliably in a two party 
situation.

I was going to go on and show other statements, more specifically by 
Bouricius, where he has made these spurious majority claims, without 
explaining what he means. That's typical for FairVote, for they *do 
not want to explain it,* they don't want the relatively ignorant to 
realize the details about ballot exhaustion and what it means. They 
want them to assume that they can get their desired majority without 
spending any more effort for it, and they absolutely don't mind that 
ignorance about this serves their cause.

We've been bringing this up for some time now, and there has been no 
change in the propaganda. It continues, because, if nobody brings 
these things before the electorate in a referendum or Ballot 
Question, *it works.* It's time for serious opposition to IRV to come 
forward, to make cogent and clear arguments against the method.

Originally, I thought that IRV was better than Plurality; as such, I 
would have been reluctant to oppose it. However, what is *actually* 
happening is that IRV is mostly replacing, not Plurality, but Top Two 
Runoff, which is a *much* better method, proably worth the cost, even 
though it, itself, suffers from one of the IRV problems: Center 
Squeeze. It's actually easy to fix, once we start making decisions 
about these things with some sanity. With a better primary method, 
roughly half of the runoffs, probably, could be eliminated (IRV may 
eliminate as many as a third of them, legitimately, the rest involve 
faux majorities), and with a better runoff method, such that voters 
can write-in candidates and still not create a spoiler effect (that's 
*very* easy to do), one has, actually, a highly democratic system 
that will *almost always* find a majority; where it doesn't, a direct 
face-off would *probably* come up with the same result, and seriously 
poor results *will not happen*. Which they *can* with IRV, though it 
is rarely a terrible result.

But if you must have a single ballot, IRV isn't the right choice, at 
all, there are very few people who understand voting systems who 
think so, and among those who support IRV, it is often because, 
supposedly, IRV has "momentum," and they don't seem to realize the 
damage that is being done by removing majority requirements.

As to "momentum," I remind readers that, by 1918, 55 towns and cities 
had implemented Preferential Voting, which is far, far more than what 
FairVote has accomplished to date with IRV. (Almost all of it was 
American Preferential Voting, i.e., Bucklin, a little was IRV.) What 
happened? FairVote gives us a brief analysis that is almost certainly 
dead wrong. My tentative hypothesis, though, is that it was oversold 
as making runoffs unnecessary, i.e., "Find a majority with a single 
ballot!" That should make us worried that, by promoting IRV, or by 
allowing it to be ignorantly promoted, we may be setting up a 
backlash, just as happened with Bucklin eighty years ago. It could 
set the cause of voting systems reform back for the better part of 
another century.

Imagine how different it would be if, instead of replacing Bucklin 
(and IRV) with top two runoff -- which was often done, I think, 
ultimately: San Francisco apparently implemented preferential voting 
in 1916! and ended up with top two runoff, but I don't know the 
intermediate history -- a majority requirement had been added back, 
so the method becomes a way to avoid *some* runoffs. In addition, if 
something as simple as Open Voting (Approval Voting) had been used 
for the runoff, or even better, a two-rank Bucklin, the runoff, with 
two clear major candidates, has a spoiler effect fix that allows 
write-in votes without a spoiler effect. This becomes almost 
completely what Robert's Rules of Order would want for straight, 
direct-democracy elections, differences would be rare. In the rare 
case that the best candidate isn't in the top two, it can be fixed! 
-- But the voters can still add a security vote for one of the 
candidates on the ballot, as a second preference.







More information about the Election-Methods mailing list