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

[Abd ul-Rahman Lomax]

At 08:02 AM 12/23/2008, James Gilmour wrote:
>Dave, I never said that I would find that result 
>objectionable.  What I did say was that I thought such a result would be
>POLITICALLY unacceptable to the ELECTORS   -  certainly in the UK, 
>and perhaps also in the USA as there are SOME similarities in the
>political culture.  It goes almost without saying that such a result 
>would be politically unacceptable to the two main parties I had
>in mind.

The main parties don't like losing, is what this boils down to. 
However, it's unlikely that any advanced voting system is going to 
magically award victories to minor party candidates more than rarely, 
at least at first. By the time it does, it will have been 
well-established as fairer than Plurality.

Preferential voting in the U.S. -- usually Bucklin -- won many 
judicial victories, definitely, losers tried to overturn it, only in 
Minnesota was there the idiosyncratic Brown v. Smallwood decision 
that did it because of the method itself. In Oklahoma, it was 
mandatory ranking.

>Political acceptability is extremely important if you want to 
>achieve practical reform of the voting system.  The Electoral Reform
>Society has been campaigning for such reform for more that 100 years 
>(since 1884), but it has still not achieved it main objective
>-  to reform the FPTP voting system used to elected MPs to the UK 
>House of Commons.

Sure. What this means is simple: if the status quo gives inequitable 
power to some, those, who, by definition, have excess power, will 
resist reform toward equity. Could it be, however, that the ERS has 
been pushing the wrong methods? Asset Voting was invented in England, 
over 120 years ago, as a tweak on IRV. It would be a far better 
method than standard preferential voting, allowing voters who only 
want to rank one candidate to vote, and it could produce true 
proportional representation with minimal compromise.

>   The obstacles to that reform are not to do with
>theoretical or technical aspects of the voting systems  -  they are 
>simply political.

Yes. And "political" doesn't mean massive voter outcry against fair 
election results. Voters don't massively reject results in the U.S. 
even when they are patently unfair, just look at Presidential 2000. 
The fact is, though, that the 2000 election was close. A close 
election is, in my view, an *inherently* poor result unless there is 
truly low preference strength involved.

>   It was for political reasons that the Hansard
>Society's Commission on Electoral Reform came up with its dreadful 
>version of MMP in 1976 and for political reasons that the Jenkins
>Commission proposed the equally dreadful AV+ in 1998.  Jenkins' AV+ 
>was a (slight) move towards PR, but it was deliberately designed
>so that the two main parties would be over-represented in relation 
>to their shares of the votes and that one or other of two main
>parties would have a manufactured majority of the seats so that it 
>could form a single-party majority government even though it had
>only a minority of the votes.

Sure. Politics. And this is why I believe that true reform must start 
with organization outside of government. The fact is that if the 
electorate were organized, it could ensure that the best possible 
candidate was on the ballot, and close elections, even with 
plurality, would be rare. In small jurisdictions, where people know 
each other and know the candidates, it's common for elections to be unopposed.

>It is sometimes possible to marginalise the politicians and the 
>political parties in a campaign if you can mobilise enough of the
>ordinary electors to express a view, but our experience in the UK is 
>that constitutional reform and reform of the voting system are
>very rarely issues on which ordinary electors will "take up arms" 
>(metaphorically, of course).

That's right. However, Bucklin was very popular in the U.S., that's 
what I'm finding. What I *don't* know is how the reform disappeared. 
It seems it was usually replaced with top two runoff, and that may 
indeed be an improvement. But holding a runoff with a Bucklin primary 
would have been even better, and about half the runoffs would 
probably be avoided.

> > In Election 2 Condorcet awarded the win to M.  Who has any
> > business objecting?
> >       52 of 100 prefer M over D
> >       53 of 100 prefer M over R
> >       Neither R nor D got a majority of the votes.
>
>Leaving aside the debate about the meaning of "majority", it is 
>clear to me that M is the Condorcet winner  -  no question.  But, as
>explained above, it is MY view that such an outcome would not be 
>acceptable to our electors.  I base my view of UK electors' likely
>reaction on nearly five decades of campaigning for practical reform 
>of the voting systems we use in our public elections.

It's hard to argue with experience, except that it's obvious that 
this experience doesn't include actual experience with such an 
outcome. James is extrapolating from other opinions that he's seen, 
just as he has with the Later No Harm criterion. It appears from what 
I've seen that, though I'd agree that voters are very likely to 
express strong support for LNH, that, in practice, with Bucklin, they 
will add the additional votes pretty much independently of LNH. I'd 
expect a stronger effect with Approval, but Approval, applied 
immediately, wouldn't suffer from *practical* harm from this, because 
the Approval votes that matter would be those where the voter prefers 
a minor candidate and the voter knows the candidate is not going to 
win. Supporters of frontrunners aren't normally going to add an 
Approval vote for the other frontrunner.

So, first of all, the 5% scenario is highly unlikely. Secondly, the 
election shown is remarkable in that, while the M voters truncated, 
which means that they had strong preference, the D and R voters did 
not truncate, which means that their preference was weak. I'd venture 
an educated guess that if the voter utilities that would explain 
these voters were studied, M would be the best winner by far. In a 
Range election with sincere votes, it would be a landslide. And 
nobody would complain, they'd have gotten a good outcome to the 
election. That's what democratic compromise is about! The magic 
compromise, the ideal, is one that is pleasing to nearly everyone, 
that minimizes serious dissatisfaction and maximizes overall 
satisfaction, which M appears to do in the scenario described.

> > As to my  "no first preferences" example, surest way to cause
> > such is to be unable to respond to them.
>
>I'm not sure what this statement is really mean to say..
>
>I understand that a Condorcet winner could, indeed, have no first 
>preferences at all.  But in political terms, such a possibility is
>not just unacceptable, it's a complete non-starter.

It's just plain impossible, so stating whether it would acceptable or 
not is shallow. Let me put it this way: if there were an election 
environment such that such an extreme vote were possible, it might 
also be such an unusual environment that the outcome would be 
extremely acceptable. What's being done is to posit some very unusual 
outcome, but then apply existing standards to it, standards based on 
systems where the outcome is doubly impossible.

And then, what is pernicious is to pretend that this relates to 
actual reforms and actual political possibilities. I've been 
promoting as the simplest, no-brainer reform, Count All the Votes. 
That's the simplest form of alternative vote, it is simply 
simultaneous consideration of the alternative votes, with a 
simultaneous abstention from participating in the pairwise elections 
between the alternative choices.

But Bucklin starts out with the first preference votes, so the first 
round satisfies the clear desire of voters to vote for their 
favorite. Then, if no majority is found, the additional votes are 
added in, one rank at a time. My opinion is that this is sufficient 
to satisfy voter's desire to specially support their favorite. That, 
by itself, doesn't mean that they will add additional preference 
votes, just as it isn't guaranteed, by any means, in IRV, and, it 
appears, many voters didn't and don't.

Bucklin is not Condorcet compliant, but will usually detect a 
Condorcet winner. Unless everyone truncates. Which is highly 
unlikely. Of course, if everyone truncates, every method behaves like 
Plurality.