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

[Abd ul-Rahman Lomax]

At 07:51 PM 12/28/2008, James Gilmour wrote:
>  All my campaigning has been to get STV-PR used instead
>of (mainly) FPTP/SMD and more recently instead of MMP.

By the way, multiwinner STV is a *far* better system than 
single-winner IRV. Later-No-Harm makes must more sense when applied 
to *representation.* It doesn't make so much sense for single-winner 
offices, where compromise is necessary (which is also that *last 
seat* being elected in a multiwinner STV election.)

(And, in fact, the nearly-ideal Asset Voting tweak on STV makes 
"Later" pretty much unnecessary! If the method remains STV, fine, it 
means voters can control the process to a degree, but, frankly, I 
doubt I would bother. Vote for the one I want to represent me, and 
this one represents me, either in the parliament or assembly, or 
represents me in the process of choosing who will represent me.)

>   My interpretation (and it is my interpretation) of UK electors' likely
>reaction to different voting systems is based on a mixture of 
>comments made in face-to-face meetings with ordinary electors, party
>activists and elected politicians; daily reading of the political 
>press and readers' letters and blogs; comments made by political
>parties (which have obvious vested interested, both pro- and 
>anti-reform); comments made by other political pressure groups, from
>trade unions, commerce, media moguls and "big money", all of whom 
>have vested interests which they try to make less obvious, mostly
>anti-reform.  And from time to time there have been public opinion 
>polls in which relevant questions have been asked, usually
>without a great deal of context and without any discussion.

The problem is that knee-jerk reactions to many of the involved 
issues can be very far from what would be a response of someone who 
has become educated on the subject. We have seen where it appears 
that FairVote advocates, if they were not outright lying, failed to 
understand the issues of majority vote, and most opponents of IRV 
here likewise missed it; yet it strikes to the core of the claims being made.

I just read again where a FairVote presentation to Los Angeles 
continues to claim that IRV guarantees a majority result. It's said 
over and over again, but it is *never* mentioned that this isn't what 
is ordinarily meant by a majority, and it is not what one gets with 
real runoff voting. It's a faux majority, obtained by simply 
disregarding all *legitimate ballots* that don't contain a vote for 
the frontrunners.

Now, if the methods being advocated allowed full ranking, some 
semblance of argument might be made that by not fully ranking, voters 
were forfeiting their right to participate in the "runoff." Except, 
of course, that this requires that voters (1) have the necessary 
knowledge to deeply rank the candidates and (2) could *tolerate* 
voting for a candidate whom they detest. Real runoffs allow the voter 
to reconsider, for one thing, and sometimes voters can still cast a 
write-in vote, and sometimes these actually win.

But the methods don't allow full ranking. Not here in the U.S. in 
major elections. 3 ranks, period. So someone who prefers three 
candidates to any of the frontrunners: doesn't count.

As I've pointed out, this could be done with Plurality: don't vote 
for one of the frontrunners, too bad, your vote doesn't count, and 
won't be considered part of the basis for "majority." Instant Runoff 
Plurality. Just take the top two candidates, set aside all other 
candidates, and notice which one has a majority of the remaining 
ballots. It's pretty much what IRV does, anyway, in nonpartisan 
elections, if anyone would bother to notice. Vote transfers only 
rarely affect the overall social ordering that results from IRV, in 
nonpartisan elections, when voters don't have that party marker to guide them.

>You can of course dismiss my interpretation of that accumulated 
>evidence, especially as that evidence is of the "grey" or "soft"
>variety and cannot be subject to the normal rigorous scrutiny which 
>one would apply to "hard" evidence.  You can also dismiss the
>evidence as being obtained from a community that for many decades 
>has been exposed to nothing but (bad) plurality voting systems and
>has accepted the political outcomes without any serious 
>protest.  (All the recent reforms of voting systems in the UK, except for a
>few mayoral elections in England, have been to introduce PR - in 
>three varieties! - but that has happened only in the last decade.)
>
>So that "plurality mindset" (for the sake of having a shorthand 
>term) is the reality we have to confront when we campaign for
>practical voting reform.  I don't need any persuading about the 
>potential merit of a Condorcet winner over an IRV winner when they
>are not the same (though there are some unresolved technical issues 
>about breaking Condorcet cycles).  I have said I think I could
>sell Condorcet to our "plurality minded" electors when the likely 
>outcome would be a strong third-placed Condorcet winner, and see
>off the vested interests that opposed reform.  But if the likely 
>outcome was a weak Condorcet winner, I am quite convinced that the
>forces of reaction would have no problem in winning the public and 
>political debate, and the reform would never happen  - or if it
>had happened, it would be reversed.

You might notice that I'm not advocating Condorcet methods, as such. 
I've become far more interested in majority rule, and how to foster 
it. A Condorcet violation is offensive to majority rule; but there 
are ways to resolve this. Real Condorcet cycles don't persist, there 
are no Condorcet cycles in social utility, when the considerations 
have become broad enough.

>We do not have in the UK a really powerful, high profile political 
>office to which the incumbent is directly elected.

And that's a good thing. We do. And it basically is an elected King. 
Our history brought us here: remarkable, isn't it? We rebelled 
against the British King, but adopted the system, only, of course, 
with some checks and balances, but still the same idea: very strong 
executive, not immediately accountable, unless he's so foolish as to 
fool around with an intern....

Geez, a President misleads the nation into a war, leading to the 
death of easily a million people, and nobody says "Boo!" Well, not 
nobody, but nobody who counts, until the next election. No, not the 
next election, the next election after that!

>  But just
>suppose for a moment that we had direct elections for the Prime 
>Minister, but within our parliamentary system.  The public opinion
>polls show support for the three main parties has fluctuated quite a 
>bit during the past year, but one recent set of figures was
>Conservatives 47%, Labour 41%, Liberal Democrats 12% (after removing 
>the "Don't knows").  Now suppose these were the voting figures
>in a direct election for Prime Minister.  The Liberal Democrat would 
>be the second choice of most Conservatives and most Labour
>voters; the Liberal Democrats' second choice would mainly be Labour 
>with some Conservative.
>
>The Liberal Democrat would be the Condorcet winner, but the 
>political consequences, both in Parliament and in the country would be
>horrific  -  given the reality of our politics.

"Horrific"? First of all, this would depend on the preference 
strengths. That someone is most people's second choice means nothing 
if we don't know the preference strength involved. "Second choice" 
could be *awful.* This is why mandatory ranking can be a Bad Idea. A 
vote for a candidate should be just that, a vote *for* a candidate. 
While it's possible to collect and use data about whether or not I 
prefer Mr. Awful to Mr. Worse, it's not what most of a voting system 
should be based on.

I.e., if we collect all the "preferences" without knowing and using 
preference strengths, we can come up with, yes, a disastrous result. 
Your "weak Condorcet winner." A good voting system *must* sometimes 
violate the Condorcet criterion!

But a Condorcet winner still *means something.* We tend to assume 
fixed preferences, i.e., the voters have voted these preferences, and 
we are going to use them. But there is ranking that is clear and 
ranking that is not, plus truncation, which leaves candidates, 
perhaps many candidates, unranked.

With truncation or weak ranking, where the voter really doesn't care 
-- perhaps because the voter hasn't considered the pairs involved -- 
Condorcet methods can come up with garbage as a result. Well, 
probably not the worst garbage, but the decision could be pretty bad.

This is why I don't like Plurality methods. Condorcet is a Plurality 
method, as usually proposed: it can elect a candidate with only a 
plurality of voters expressing a preference for that candidate. 
That's a weak winner! It doesn't matter how much shuffling of votes 
one did to get there: either a majority of voters supported and 
"approved" the candidate, or they didn't.

Some ranked or rating methods don't specify approval, and, then, the 
ballot can't be used to test for majority acceptance of the result. 
That's a problem! Open Voting, or Approval, of course, has no problem 
with this, but Range does, unless some approval cutoff is specified.

I'm saying that two round methods are almost inherently superior to 
single round methods, other things being equal. There is a *huge* 
advantage to that second round, and improving two round methods 
becomes a much simpler necessity to get the best possible winners 
into the runoff, instead of insisting that everything be discovered 
in a single step. A plurality primary is primitive and quirky, but if 
it finds a majority, and it usually does in a two party system! or in 
many nonpartisan elections when there aren't insane numbers of 
candidates, it can be good enough.

STV, though, does its first elections without much opposition (under 
current conditions). My guess is that the bulk of voters get the 
representation they want. It would be interesting to get some numbers 
on this! Under Asset, we would see fewer direct elections (quota of 
votes from first preferences), but also fewer wasted votes and fewer 
compromises with significant loss of value. The "compromises" made in 
Asset may not be losses at all, they may be gains. I.e., I may gain a 
better representative than I would have known to vote for!

STV is certainly respectable! But it could also be improved, in a 
really spectacular way that could possibly evolve into something very 
much like direct democracy, the best of direct and representative 
democracy. The actual voting system could remain STV.... with 
Carroll's simple suggestion for handling exhausted ballots.




>   You can blame that on the "plurality mindset" of our electors and 
> of all the other
>political stakeholders if you like, but that is not going to change 
>the political reality.  Now that is the reason why I see
>problems with a weak Condorcet winner.  And because such an outcome 
>is a likely outcome, that is why I see problems with
>recommending Condorcet voting for such single-office elections.
>
>It must be for others to judge the extent to which any of this is 
>interpretation is relevant to their own countries, but from what I
>read about politics in countries that were cursed with the British 
>legacy of FPTP/SMD voting systems, I fear they too would likely
>face similar problems with a weak Condorcet winner.
>
><snip>
>
>
> > IRV is equally vulnerable. Fear of change and a misunderstanding of
> > one-person-one-vote work against us both (although I don't know if
> > one-person-one-vote is treated as a quasi-Constitutional
> > principle in the UK).
>
>I think there are three issues here.  First, any reform that breaks 
>or reduces the power of the current vested interests will be
>opposed, but that opposition can be overcome.  Second, 
>one-person-one-vote is not written into our constitution (we don't have
>one!), but it is the assumed basis for all our many voting systems, 
>even for the pernicious multi-winner FPTP "block vote".  Third,
>IRV has a "political" weakness that is the obverse of the weak 
>Condorcet winner  -  IRV fails to recognise and elect the strong
>third-placed Condorcet winner.  This defect in IRV is recognised in 
>the UK  -  the elimination of the candidate "who is everyone's
>second choice".  But that failure of IRV is accepted political by 
>our electors and our political community, I suspect because it
>fits with the "plurality mindset" and it protects them from the weak 
>Condorcet winner.  I have not heard it expressed in these terms
>in the UK, but that should be no surprise as Condorcet voting does 
>not feature at all in any public debate about voting reform in
>the UK.
>
>James
>
>No virus found in this outgoing message.
>Checked by AVG - http://www.avg.com
>Version: 8.0.176 / Virus Database: 270.10.0/1866 - Release Date: 
>27/12/2008 20:49
>
>
>----
>Election-Methods mailing list - see http://electorama.com/em for list info