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

[Abd ul-Rahman Lomax]

At 03:17 AM 12/4/2008, Kristofer Munsterhjelm wrote:
>James Gilmour wrote:
>>Abd ul-Rahman Lomax  > Sent: Wednesday, December 03, 2008 10:52 PM
>>>The tragedy is that IRV is replacing Top Two Runoff, an older
>>>reform that actually works better than IRV.
>>I have seen statements like this quite a few times, and they puzzle
>>me.  I can see the benefit in TTRO in knowing before voting at the
>>second stage which two candidates will actually be involved in the
>>run-off.  But what concerns me is the potential chaos in getting to
>>that stage.  The French Presidential election of 2002 is a good
>>example of the very bad results that can come from the first round of
>>TTRO.  And we have seen similar problems in some of the mayoral
>>elections in England where the so-called Supplementary Vote is used
>>in which the voters can mark their first and second preferences but
>>only the second preferences for the first stage Top-Two candidates
>>are counted.  In such circumstances the outcome from TTRO is very bad
>>and I should have thought that an IRV election would have given a
>>much more representative result.  Condorcet might be better still,
>>but that's a different debate.
>
>I'm not Abd, but I think the argument goes like this: in TTR, if a 
>(usually) third candidate gets enough FPP votes to make it to the 
>second round, that candidate has a real chance of winning, since the 
>second round will be focused on those two candidates alone, whereas, 
>on the other hand, if it's IRV, then IRV's chaos may deprive the 
>candidate of its rightful victory, and even if it wouldn't, people 
>can only vote for the third candidate that would become the winner 
>as one of many, not as one of two.
>
>If that's right, then the Supplementary vote should give 
>significantly worse results than TTR, simply because people can't 
>discuss and realign between the first and second rounds.

For the benefit of those who aren't familiar with the terminology, 
"Supplementary Vote" is top-two batch-elimination IRV. In the United 
States, there are or have been a few implementations of SV, and 
FairVote claims these as IRV successes.

IRV with sequential bottom elimination is probably better than batch 
elimination, because there is, at least, an opportunity for vote 
transfers from other eliminated candidate. However, it is far more 
difficult to count and what I've seen with U.S. IRV in nonpartisan 
elections, the vote transfers tend to not change the positions of the 
top three. If we imagine only three candidates, which has been the 
case with some IRV elections, this is identical to Supplementary  Vote.

For whatever reason, the fact is that in TTR in U.S. nonpartisan 
elections (or in party primaries which are effectively nonpartisan in 
this sense), we see "comeback" elections where the runner-up in the 
primaries wins the runoff, roughly one-third of runoffs. This hasn't 
happened yet in these IRV elections in the U.S.

Now, I have not studied the most recent elections. One writer 
responded that the Pierce County, Washington, elections showed an 
exception. If I'm correct, that was a partisan election. When there 
are minor parties present, with relatively predictable vote 
transfers, (such as, usually, Green to Democrat), IRV corrects for 
the spoiler effect. *However*, don't celebrate!

If the use of a runoff method encourages more parties to be on the 
ballot, majority failure becomes common. That is what had happened 
with San Francisco, in the nonpartisan elections where they were 
using top-two runoff. That will continue with IRV, and it has. 
Usually a majority is not found through transfers if it was not found 
in the primary, and, in fact, usually the first rank choices 
generally express the results after transfers, as far as overall rank order.

This probably makes the world safe for major parties! A third party 
is prevented from spoiling elections, which removes some of their 
power. Again, for whatever reason, IRV is associated with strong 
two-party systems, whereas top-two runoff, around the world, is 
associated with vigorous multiparty systems. Both methods display the 
problem Mr. Gilmour is concerned about, it's odd that he pins it on 
TTR and not on IRV, which is merely *maybe* a *little* better in that respect.

We may speculate that with IRV, Le Pen would have slipped behind 
Jospin, thus putting the true top two if second place preferences are 
considered into the runoff. To come up with a guess about that would 
take more information than I have and effort than I can spare. The 
point here is that it can happen, and does happen, that Center 
Squeeze pops up, it is not a rare effect. Center Squeeze is certainly 
a problem with TTR, but IRV isn't the fix.

IRV fixes the spoiler effect in Plurality, but so does TTR, and so 
would other voting reforms, starting with the terminally simple Open 
Voting (Approval). Bucklin uses the same three-rank ballot as RCV 
(IRV), but allows far more flexibility on the part of voters, who can 
use the three ranks to vote for many more than three candidates.

Example. There are twenty candidates, San Francisco has more than 
that on the ballot. The voter has a favorite and a second favorite, 
but is worried that these are not going to make it to the final 
round, and the voter really dislikes a frontrunner. With IRV, the 
voter needs to know, to cast an effective vote, who the frontrunners 
are, and to make sure that the preferred ones are supported. That can 
be quite difficult when there are twenty candidates! Perhaps as a 
result, one race in San Francisco, with over twenty candidates, was 
won with less than 40% of the vote, there were large numbers of 
exhausted ballots.

With a Bucklin ballot, even the Duluth form, where voters could add 
multiple approvals only in the third rank, the voter could vote for 
sincere favorites in first and second rank, and then for every 
candidate but the worst frontrunner -- and also not for anyone worse, 
if the voter has such an opinion. With RCV, the voter would have to 
guess who else might be in the running in the last round. In France, 
they would have guessed wrong, that was the "chaos." Few expected Le 
Pen to edge out Jospin. (By less than one percent.)

Allow equal ranking at all ranks and Bucklin becomes easier to vote 
than IRV. Very simple and effective strategy:

If you have a reasonable preference, then vote first and second ranks 
as unique candidates. If not, then equal rank. You don't have to 
choose, you can equally support candidates where you like both and 
don't have a strong preference. Then, third rank, you can support 
anyone whom you'd rather see win the election than have it go into a 
runoff (if a majority continues to be required, which it should). And 
if you have a strong preference against one of the frontrunners, you 
can apply maximum alternative vote strength by voting for every 
candidate better than this one. That's standard Approval strategy, 
but, here, it's easy even with three or more candidates.

Bucklin. It deserves much more attention. It was popular. It worked. 
It's cheap to count. No eliminations, so no Center Squeeze (unless 
everyone bullet votes, but that's very unlikely; majority failure is 
more likely in a highly partisan environment, where every voter is 
fiercely loyal to their party or candidate and imagines they'd rather 
die than see their vote "help" someone else. But we don't need to 
drive a stake through the heart of bottom-ranked candidates in order 
to satisfy these voters; their lower ranked votes will only help 
another candidate beat their favorite, as some would rant and rave 
about, if no majority is found and its necessary to start 
compromising. Compromising means that you may not get your favorite! 
What do we think about people who insist on No Compromise!

Probably much the same as the reviewer of Woodall's original paper 
that defined Later No Harm: as to the criterion, disgusted.

In fact, it's not true that an additional vote "helps another hurt 
the favorite." It's really an abstention in that candidate pair, a 
"stand-aside." It does not act to help the other candidate more than 
the favorite; rather it equally helps them.

Note that if we require a true majority, and don't coerce votes (as 
in requiring full ranking), no single-ballot method satisfies Later 
No Harm. It is incompatible with the requirement that a majority vote 
for the winner. "Instant runoff" does not simulate real runoff except 
under the artificial constraints of imagining that voters don't 
change preferences, the same voters vote, and that voting Bush > 
Hitler is a vote for Bush.

The solution to Center Squeeze, which is what happened in France in 
2002, isn't IRV, it's using a better first round method, one not 
susceptible to Center Squeeze. IRV is very much the wrong choice 
there. Approval would work better, no cost. Bucklin would work even 
better, probably, using a practical preferential ballot with easy 
counting. And no Center Squeeze, almost certainly.