[EM] Why I think IRV isn't a serious alternative 2

Abd ul-Rahman Lomax abd at lomaxdesign.com
Sun Dec 28 16:30:10 PST 2008


At 12:02 PM 12/28/2008, Terry Bouricius wrote:

>I don't want to re-hash our Wikipedia argument about whether an exhausted
>ballot in a full-ranking-possible IRV election should be treated as an
>abstention -- like a stay-home voter in a runoff (excluded from the
>denominator), or treated as a vote for none-of-the-above (included in the
>denominator), in deciding whether the winner has "a majority." I will
>simply say, that I have not resorted to "deception," in any of my writing
>about IRV and majority winners.

I'm not aware that I've attributed this personally to Terry. However, 
the description above shows a certain slipperiness. The term 
"majority" as applied to elections has some very well-established 
meanings. If we say that a candidate got a majority in an election, 
we mean that a majority of those voting supported that candidate. 
There are quibbles around the edges. What about ballots with marks on 
them but the clerk can't figure out what the marks mean? Robert's 
Rules are clear: that's a vote, part of the basis for a majority. 
However, public election rules generally only admit, as part of the 
basis for a majority, ballots containing a vote for an eligible candidate.

By making up some complicated definitions and conditions, Terry can 
attempt to confuse the issue, which is what this is looking like. The 
San Jose initiative ballot arguments, 1998, are worth looking at:

http://www.smartvoter.org/1998nov/ca/scl/meas/F/

>Impartial Analysis from the County Counsel
...If a candidate now has a majority of the ballots, that candidate 
is elected. If not this process is repeated until one candidate 
receives a majority of the ballots. IRV eliminates the need for the 
second, separate, runoff election. ...

The analysis mentions candidate eliminations. It does not mention 
ballot eliminations, and that is what is necessary to consider the 
IRV "majority" a "majority of the ballots." Terry, it is sophistry to 
claim that the ordinary person would not understand "majority of the 
ballots" would mean "the majority of all ballots legitimately cast in 
the election."

Note that Terry here refers to "full ranking possible" IRV elections. 
It's *possible* to make that argument if, indeed, full ranking is 
possible. That has not been a proposal in San Francisco. Further, 
with real runoffs, the voter makes a choice to vote or not. In the 
primary, the voter may not even have known the relevant candidates 
well enough to make a choice.

But the point about deception, here, is that the arguments were made, 
very plainly, that a majority was still going to be required, and any 
ordinary person, unfamiliar with what actually happens with real IRV 
elections (massive truncation), would read it as just that, a 
majority of the ballots cast. Not only of ballots containing votes 
for uneliminated candidates. As I've pointed out, with that kind of 
sophistry, one could claim a majority in any Plurality election: just 
do *exactly* the same thing: eliminate all but the top two.

Now, the above was from the County Counsel, supposedly neutral. I 
have no reason to doubt the neutrality, in intention, but the Counsel 
swallowed a bit of propaganda! What is the "need for a second, 
separate runoff election"? It is nothing but the need to find a 
majority, a real majority of ballots cast in an election where all 
registered voters may vote. Does IRV satisfy that need? It does not. 
Robert's Rules knows that, and you should know it by now.

What you are doing is trying to deflect this with arguments about the 
numbers of voters in each election. But these are separate elections, 
and there is no fixed relationship. Sometimes runoffs have more votes 
than the primary. With the real Ranked Choice Voting that voters are 
getting, they *cannot* fully rank, so this is simply nonsense. A 
voter may easily vote sincerely using all the available ranks, but 
still the voter is excluded from the basis for a majority. Literally, 
their vote doesn't count, and they have no opportunity to remedy that 
in a runoff, as they would under top two runoff.

The arguments went on....

Steve Chessin et al wrote, in the pro argument:

>IRV reduces the cost of campaigns. Candidates only have to raise 
>funds once, and the county only has to pay for one election. Most 
>elections for open seats have resulted in runoffs - take the current 
>Sheriff's race as an example. With IRV the public gets a majority 
>winner with only one trip to the polls.

"Majority winner?" What's that? The word has an ordinary meaning. In 
Australia, with Preferential Voting, they use the term "Absolute 
majority." That's what is required to win. But: they require full 
ranking, a ballot which is not fully ranked is spoiled, not counted. 
So they get a majority. In Queensland and some other places where 
they have Optional Preferential Voting, they change the wording, it 
becomes a majority of ballots containing a vote for a continuing 
candidate. Sure, you can claim that, "Well, that's what we meant!" 
But you know full well that, politically, if Chessin had said that, 
the San Jose initiative -- the first success in the FairVote campaign 
-- would have been dead meat. It was close enough anyway.

A Libertarian smelled a rat; he did realize the problem (unlike any 
of the opponents in San Francisco, a very remarkable lacuna) but he 
also was confused:

(The Rebuttal to the Pro argument, a piece of which was quoted above)

>Vote NO on MEASURE F.
>
>LET'S NOT BUY A PIG IN A POKE. Let's not leave it to anyone but YOU 
>the taxpayer to decide on such a potentially expensive and 
>revolutionary method of determining who gets control of government, 
>and the power to tax and spend.
>
>It is UNCLEAR; it will mean whatever the Supervisors want it to 
>mean. IT WILL BE OUT OF YOUR CONTROL.
>
>THIS IS ONE OF A SERIES OF MEASURES TO TAKE YOUR VOTING RIGHTS FROM 
>YOU AND GIVE IT TO A BOARD, THE SAME BOARD THAT WANTS TO INCREASE 
>THEIR TERM OF OFFICE! PROTECT

The Con argument by this party was:

>A Triumph of Jargon and Nonsense over Substance and Common Sense!
>
>This confusingly written measure insults the intelligence of all 
>voters in Santa Clara County. The vague language seems to say that a 
>candidate earning less than 50% of the total votes cast could be 
>awarded the office. Does this measure mean that the Board of 
>Supervisors picks the winner of an election regardless of the vote?
>
>In the American form of government, runoff elections force 
>candidates to convince a majority of voters that they are fit to hold office.
>
>Tell the Supervisors that taxpayers consider it money well spent to 
>hold an election to determine clearly the best-qualified candidate for office.
>
>Protect your voting rights! Vote No on Measure F.
>
>KATHLEEN O'CONNELL-SUNDARAM, President,
>Santa Clara County Taxpayers Association
>
>MARVIN B. RUDIN, Chairman, Santa Clara County Libertarian Party

I'd guess that these critics didn't really understand preferential 
voting, unfortunately. The Board of Supervisors didn't have anything 
to do with it.... except that they would set the details of the 
method, if I've got it correctly. It's even possible that they could 
continue to require a majority. Implementation details were not 
specified; however, since the measure does also clearly claim that 
the method makes runoffs unnecessary, the simplest interpretation 
would be that a true majority would no longer be required. But, of 
course, that contradicts the claims about majority. I.e., Rudin was 
right. The measure was confusing.

It gets more specific.

In response to the Con argument, Chessin et al wrote:

>The fears of the opponents are unfounded.
>
>Measure F does NOT repeal the majority requirement for electing 
>county officers. A candidate still must earn more than 50% of the 
>total votes cast in order to be elected; it just allows the majority 
>winner to be determined in one election by using the more modern 
>technique of Instant Runoff Voting.

This is absolutely untrue, there isn't any wiggle room. I'll also 
note that the method isn't "modern." It's called Ware's method, 
sometimes, or English Preferential Voting, it's very old, back to the 
19th Century. Bucklin, American Preferential Voting, as it was known, 
is substantially newer. You want "modern," go for Range Voting. But 
no voting system can determine a majority winner, guaranteed, in one 
election. Bucklin was sold with the same argument and, we might note, 
Bucklin disappeared. I still don't know why, Later No Harm probably 
had little or nothing to do with it. What might have been more 
relevant could be that it was oversold, and thus all these arguments, 
being repeated in jurisdiction after jurisdiction, that IRV 
guarantees majority winners, could easily backfire, torpedoing the 
whole voting system reform effort. It all happened before, Terry.

In this rebuttal, Chessin apparently recruited the chair of the 
California Libertarian Party, who endorsed the measure.... I was 
pretty surprised to see a personal attack on the one presenting a 
ballot argument -- a local Libertarian --, but I suppose it's done.

>By the way, the Libertarian Party of Santa Clara County wants you to 
>know that they did not authorize their Chair to sign the argument 
>against. In fact, many Libertarians support IRV, as do many Greens, 
>many Democrats, many Republicans, and many Independents. IRV does 
>not favor one party over another, it just eliminates the need for a 
>second separate election while still preserving majority rule.

IRV doesn't preserve majority rule. It's a plurality method. In real 
elections, in the U.S., where a majority has not been found in the 
first round, it almost never finds a majority. This seems to have 
been overlooked, not even Robert's Rules seems to be aware of it, for 
they suggest sequential elimination preferential voting in an attempt 
to find a majority, when, it appears, at least in the nonpartisan 
public elections which would most closely resemble private 
organizational elections, it is very inefficient at that. Bucklin is 
probably more efficient, because it counts all the votes, but 
majority failure did occur with Bucklin elections, albeit at a 
possibly lower rate.


>I am, however, interested in your statement:
>  <snip>
>"It's simply a fact: top two runoff is associated with multiparty systems,
>IRV with strong two-party systems."
>
>I take it, that your use of the word "associated" means you are not
>actually claiming any causality, correct?

It means that I'm not an expert in the field, and claims of causality 
can be suspect when you have only a few examples to look at. What I 
do see is that IRV, simply from how it operates, does not give third 
parties or minor candidates a leg up, compared to top two runoff. 
With IRV, you can claw your way up to second place, you still lose. 
With top two runoff, do that and you then have a chance to convince 
the electorate, who will now vote on the specific election in front 
of them. I've read reports of where a minor candidate did just that, 
and won, and these are "comeback elections," and I found them in 
roughly one out of three real runoff elections; FairVote has reported 
29%. And no comeback elections among the nonpartisan elections being 
held in the U.S. with IRV, which is nearly all of them.

With partisan elections, vote transfers move in predictable ways, 
more or less, and thus vote transfers will tend to go preferentially 
to parties that are allied or associated in some way, such as Green 
votes to Democrats, perhaps. Or the Human Rights Party to the 
Democrat in Ann Arbor, MI.

This is the spoiler effect situation that IRV does indeed fix, 
roughly. And it is only particularly meaningful in a two-party 
system. IRV makes the system safe for the two major parties, so minor 
activity by third parties is largely moot. Top Two runoff can result 
in real upsets, sometimes, where a minor candidate *wins*.

>  Can you give examples of
>countries that use only winner-take-all Top-Two Runoffs (TTR) elections
>(and no form of PR) that has a multi-party democracy (by which I mean that
>more than two parties regularly succeed in electing candidates). It seems
>to me that the distinction you are trying to make between TTR and IRV in
>terms of multi-party democracy is specious, as both are winner-take-all
>and inevitably not conducive to multi-party democracy...What matters is
>whether the country uses a form of PR for legislative elections,
>regardless of what method is in place for electing single-seat executives.

I'm not familiar enough to do this at this point. Perhaps someone 
else is. Absolutely, proportional representation is a strong effect, 
possibly stronger. However, Australian has PR in the Senate, with 
*little* third party strength. It could be that the two-party bias of 
IRV (it is essentially the bias of Plurality; IRV sometimes shifts 
the winner between the two majors, that's all) is enough to keep the 
minor parties down.... or it could be that other effects prevail.

My main concern, here, was the deception involved in selling IRV, and 
I'm disappointed that Terry doesn't seem to recognize the depth of it.

People want majority winners. Terry wrote a desire for a majority 
into his IRV legislation in Vermont. Majority rule is a fundamental 
principle in democracy. And Top Two Runoff is the closest we've 
gotten, in the U.S., to it. When push comes to shove, what's being 
done is to sell democracy for the cost of a few runoff elections, in 
favor of an ersatz "majority" that isn't.

FairVote provides a sweet song: you can have the majority you want, 
with one ballot. Cheaper! More Convenient!

If I assume that Steve Chessin wasn't simply lying through his teeth, 
even IRV activists have been very confused about this. "Cheaper"? IRV 
can be so expensive to count that it can take a lot of saved runoffs 
to make up for the conversion costs, and it's still debatable. 
Bucklin? Practically free. Just about the same benefits, in fact, 
only more *real* majorities. But *not* a replacement for real runoff 
elections, you know that, or should know it, Robert's Rules of 
Order's editors know that, and anyone who actually studies the 
methods knows that.

Top Two Runoff is probably the most sophisticated reform to be used 
in the U.S. All it needs to be *much* better is better primary and 
runoff methods. If write-ins are allowed in the runoff, it becomes 
*almost* what Robert's Rules would prefer. And write-ins are allowed 
in runoffs in many places. Sure, it can trash the "majority." 
Priorities. Which is better, free voters or a claim of majority? If a 
majority is obtained by preventing the voters from expressing 
themselves, it's a false majority.

In fact, the absolute majorities in Australia are coerced majorities. 
It was tried here, you know, with Oklahoma Bucklin. That's why they 
didn't run Bucklin elections in Oklahoma, the voting systems 
reformers there got the bright idea that it would be easier to find a 
majority if they required all voters to add one or more additional 
ranked candidates. (If there were three or more candidates, they had 
to use the second rank, and I think that if there were four or maybe 
five, they were required to add a third rank.)

Mandatory full ranking will never fly in the U.S., it's been ruled, 
and I believe correctly, unconstitutional, in Oklahoma, I don't know 
if it was even attempted anywhere else. So what happens in Australia 
with Optional Preferential Voting? Lots of truncation. And Plurality winners.

And besides, full ranking isn't even allowed in most places using IRV here.

Now, has anyone told San Jose (the Santa Clara County Board of 
Supervisors) that there is this little problem? That the 1998 measure 
was based on a deception? Will they go ahead and remove the majority 
requirement from their election code? Or will they implement IRV with 
a majority requirement?

The latter would be a bad idea! Spend all that money, and save no 
runoffs. Bucklin would be better. Same ballot (3-rank RCV ballot). 
But additive canvassing, far simpler. Cheap. Would save more runoffs 
than IRV. Probably nearly identical voting patterns; Bucklin doesn't 
*technically* satisfy Later No Harm, but where it counts, it does. 
(Think: who needs to add additional votes? Not supporters of two 
frontrunners! They might, they might not; but the ones who support 
minor candidates are largely not concerned about later no harm; if 
they are, they will truncate, which is fine. That's why there are 
still some runoffs.)

Lewis Carroll actually saw the basic problem with the STV method back 
in 1884. (Modern my foot!). He published a solution. The problem is 
that many voters -- perhaps even most -- have insufficient knowledge 
to rank more than one or maybe two candidates. They know who their 
favorite is, presumably, which is, with Plurality, enough. 
Practically by definition, most voters support a major candidate. 
Carroll figured out how to use STV with many voters just voting for 
one, even if that one is eliminated....

No, "modern" isn't modern. If Carroll's idea were implemented, IRV 
would work far, far better. So would Plurality, since it would amount 
to almost the same thing. (I believe that if Carroll's idea were 
implemented, so many voters would vote for only one that an STV 
system would be abandoned for a vote-for-one system, except that 
there is a simply way of handling overvotes without discarding them.

It was reinvented by Mike Ossipoff and Forest Simmons in the late 
1990s, published in this very mailing list. It was reinvented by 
Warren Smith in 2004, I think it was. Ossipoff and Simmons called it 
Candidate Proxy. Smith called it Asset Voting, and Carroll used the 
same metaphor: the voting rights of an exhausted ballot, not used to 
elect anyone, become the "property" of the candidate in first 
position on that ballot, to recast at will.

So if your vote is wasted, you now have someone to blame for it! 
(Perhaps yourself, for voting for such a person!)

Suddenly, *totally* sincere voting, no compromises, becomes possible. 
There is absolutely no reason to vote "strategically" in Asset 
Voting. It would transform democracy. Maybe that's why it didn't happen!

(Bucklin was certainly an advanced system compared to everything else 
in the U.S. at the time. Except what replaced it? Sometimes, top two 
runoff. Am I correct that this also happened with IRV?)





More information about the Election-Methods mailing list