[Election-Methods] RELEASE: Instant Runoff Voting (Chris Benham)

[Abd ul-Rahman Lomax]

At 01:09 PM 6/23/2008, Stéphane Rouillon wrote:
After a nice discussion about keeping cool, 
usually a great idea if one can manage it. On the 
other hand, sometimes getting a little hot can get things done.

>So now can you acknoledge that IRV is better than FPTP ?
>I can accpet IRV being worst than any other 
>method (even if I do not agree all the time)
>but FPTP has to be worst!

I would once have agreed with this, and, still, 
under some conditions, would agree that IRV is, 
in some ways, better than Plurality. However, it 
also can be worse in some ways, and there's the 
rub: what is the balance? Political activists, 
once they've made a decision -- and they tend not 
to be the best of decision-makers, not thoroughly 
investigating before deciding -- they become like 
a guided missile that can't be recalled, it will 
continue trying to hit its target even if it 
would become obvious, to an intelligent pilot, 
that it is an Iranian Airbus rather than a military jet.

But, in the United States, where I live, IRV 
isn't replacing pure FPTP. It's replacing Top Two 
Runoff (TTR). And it is pretty clear to me that 
TTR is superior in just about every way, that 
common arguments used to claim the reverse are 
spurious, and that the one possible superiority, 
cost, becomes, actually, an argument for 
different reforms, not IRV, with its quirks and 
its typical reproduction of Plurality results in 
nonpartisan elections, and, in partisan 
elections, its reproduction of TTR problems without the ameliorating factors.

Why is IRV being pushed in the U.S. as if it were 
a positive reform? The initiative did not come 
from voting systems experts. It came from 
political activists, interested in proportional 
representation. The history is pretty clear: 
There was a conference in the early 1990s to 
consider how to bring PR to the U.S. Out of that 
conference (but not by the conference itself, it 
was a small group who acted on their own), came 
the formation of the Center for Proportional 
Representation, I think it was called. At some 
point, I'm not clear when, they changed their 
name to the Center for Voting and Democracy, to 
represent a wider focus and to match the strategy that had been developed.

Possibly the best method in common use for 
Proportional Representation is Single 
Transferable Vote. We now know how to do it 
better, to be sure, but STV is pretty good, and 
it gets better the more members districts have. 
Now, problem is, STV is a complicated voting 
system, more complicated even than IRV, because 
the vote transfers get pretty hairy as members 
are elected. (I'm assuming that one of the better 
methods is being used. STV if it is just the 
"top" candidates isn't so good, it's going to 
imitate, more or less, plurality-at-large.) This 
was considered an obstacle, and correctly so. 
There are other methods which aren't so 
difficult, that can be even more accurately 
proportional and simpler for voters, but 
remember, these were not voting systems experts 
and they didn't want to invent something new. 
(Though, in my opinion, the *best* method, and 
surely the simplest, would be Asset Voting, as 
first described by Lewis Carroll in the early 
1880s and reinvented by Mike Ossipoff, Warren 
Smith, and possibly others, recently.)

What to do? I can imagine the excitement when the 
name "Instant runoff voting" was proposed. Runoff 
voting was in used in the U.S., and it costs 
money to hold those runoff elections. The extra 
cost of STV could be justified by the cost 
savings from runoffs. So if they could get 
jurisdictions using top two runoff to establish 
IRV, it would then be a smaller step for these 
jurisdictions to move to proportional representation.

It was a political strategy, and it did not take 
into account the serious problems of IRV. For 
starters, IRV is used in two-party systems, and 
its effect is to protect the major parties from 
election spoilage by minor parties. It does allow 
third parties to exist, but makes it very 
difficult for them to actually thrive. Because of 
PR in the Senate, I think it is, they can 
exercise some power. But they don't win seats in 
the House where IRV is used. (I'm could be 
getting things mixed up, but we have Australian 
readers who, I'm sure, will correct me if I get it wrong.)

STV is a good multiwinner method, as I mentioned, 
for proportional representation, because the 
members it elects are clearly good choices, 
except for the last one. When you are electing 
many members, that the last member can be a bit 
off doesn't matter so much. But when that's the only one elected ....

Further, IRV as generally used in Australia 
(Preferential Voting) is different from IRV as 
being implemented here. First of all, full 
ranking of all candidates is required, or the 
ballot is informal and is not counted. This has 
two consequences: a majority is always found, but 
the majority can be, to some extent, coerced, and 
"donkey voting" is common, apparently, where 
voters simply make the easiest marks they can to 
fill out the ballot. (Thus they use Robson 
Rotation to randomize ballots to minimize the 
effects of this.) Secondly, PV is being used for 
partisan elections, which behave quite differently than nonpartisan elections.

In the U.S. what is being implemented is, almost 
entirely, Ranked Choice Voting with limited ranks 
(Three in San Francisco, even with as many as 22 
candidates on the ballot), and it is replacing 
Top Two Runoff, not Plurality, and TTR is 
*mostly* used for nonpartisan races (and I think 
all the IRV implementations have been nonpartisan).

Now, IRV was used before in the U.S. It was 
replaced. What was it replaced by? Isn't that an 
interesting question? Why is it that FairVote, 
the organization that the Center for Voting and 
Democracy, never mentions this in public debates 
over IRV? In some places IRV was replaced for 
clearly partisan reasons, and sometimes that gets 
mentioned. But IRV was used for party primaries 
in some places. Why was it replaced? The reason 
I've seen -- and I'm not certain that this is the 
case -- is that not enough voters were using the 
additional preferences to justify the complicated 
ballot, so it was replaced with .... ta dah!

Top Two Runoff. Which allows voters to vote for 
one only, a fairly easy choice to make, and which 
*usually*, in nonpartisan elections (and it is 
arguable that party primaries are nonpartisan) 
either makes the same choice as IRV, or a better 
one. Like IRV, it can miss a compromise winner, 
but then it makes the best choice among the 
remaining ones, and -- big secret, not widely 
known, often coming as a surprise -- voters can, 
if they want, where write-in votes are permitted 
in the runoff, *fix* an error in the top two choice. And occasionally they do.

IRV is sold as a simulation of runoff voting, 
hence the name; the implication is that it 
produces the same results, but without an 
expensive extra election. However, when we look 
at actual runoff elections, we see "comeback 
elections," where the runner-up in the primary 
ends up winning the runoff; it happens about 
one-third of the time. It's worth looking at why 
this happens, it's pretty clear to me, but I 
won't go there yet. At this point, it's enough to 
note that IRV is *not* producing these comeback 
elections, and this is quite solid from the data 
we have. Even in partisan environments, comeback 
elections can be expected to be less common than 
one out of three: the exception is where there is 
vote-splitting from an upstart third party.

So, it's possible to argue that IRV is an 
improvement over FPTP in a situation where (1) a 
majority isn't required to win, and (2) it's a 
partisan election. In nonpartisan elections, IRV 
is simply reproducing the results of Plurality, 
for much higher counting cost. Why does this happen?

Apparently, in nonpartisan elections, the voters 
for a particular candidate tend to be a 
representative sample of the entire voter 
population with regard to its opinion about 
*other candidates*. In other words, if, say, 30% 
of voters think A is best (if B is excepted), 
then 30% of voters who prefer B will have A as 
their second choice. I didn't expect to find 
this, it was quite a surprise to see it, but I'm 
not the first to notice the phenomenon, Antony 
Green of ABC notes it in his analysis of 
preferential voting in Australia. (In other 
words, it even happens with partisan elections, 
to a substantial extent, meaning that voters 
aren't as partisan as political activists would like them to be.)

Now, in some places in Australia, they have 
Optional Preferential Voting. Where they have PV, 
they call the quota an "absolute majority," 
meaning that all ballots with legal votes are 
considered. Where they make full ranking 
optional, they change the quota to a moving one, 
i.e., a majority of all ballots containing votes 
for continuing candidates. And, according to 
Antony Green, bullet voting is on the rise in 
those places. Which would eventually lead to the 
situation encountered in the U.S.: not enough 
voters using additional preferences to make the 
trouble of using STV for single-winner worth the effort.

Top-two runoff is known to favor the survival and 
occasional wins of minor parties. It is, again, 
obvious why. It is less of a leap to make it to 
second place in an election than it is to make it 
all the way to the top, so third party candidates 
manage to do it more often. And then they will be 
taken seriously, and, we might roughly predict, 
they will win perhaps a third of the time. Unless 
they are widely considered offensive or fringe. 
So Le Pen in France wasn't going to win the 
runoff, and neither was David Duke in Lousiana. 
Neither France nor Louisiana allow write-ins in 
the Runoff, (a separate issue, to be considered 
on its own), which could fix the problem that 
both elections passed over, almost certainly, the 
Condorcet winner, forcing many voters to "hold 
their noses" and vote for the lesser evil. Top 
Two Runoff gives third parties (and independent 
candidates) an opportunity to convince the public 
they deserve office. IRV tends, strongly, to 
elect the first preference leader; the vote transfers simply maintain the lead.

IRV in the primary for TTR, with runoff if there 
is majority failure, would be better than 
Plurality in the primary, I think, but that would 
defeat the rationale for the reform to some 
extent. Perhaps one-third of the time (it's been 
a bit less), IRV finds a majority winner, so it 
would avoid roughly one-third of runoffs, not 
considered worth the expense, I'd predict.

However, there is a simpler method that would 
probably find a majority winner about half the 
time, though it's a bit speculative, and that is 
Bucklin Voting. I see no sign that it would do 
worse than IRV, but it is *far* easier to count. 
Just count all the votes and add them up in stages.

And there is an even simpler method that might do 
about as well: Approval Voting. Just Count All 
the Votes, period. Stop discarding ballots with 
overvotes, count the votes. It's clear that in a 
two-party environment, it would fix the spoiler 
effect about as well as IRV, but the benefit 
continues even as third parties start to come up 
in position, whereas that is precisely when IRV breaks down.

The biggest problem with Approval, of course, is 
that it does not allow voters to cast that 
contingent vote while still remaining able to 
express first preference. That's why Bucklin may 
be more attractive, it allows it. (Indeed, as 
previously implemented, it required it. First 
preference was required to be for one candidate 
only. But I see no good reason for that. If a 
voter actually wants to be able to equally rank 
two candidates, I see utterly no reason to 
prohibit it. These are really alternative votes, 
because, in the end, *at most* one of them counts for the winner.

So, is IRV better than FPTP? Depends, doesn't it? 
IRV produces the same results, for the most part, 
as FPTP; where it does not, it is still highly 
protective of two major parties and, as what is 
really a Plurality method in disguise, it tends 
to maintain a two-party system. And the 
comparison between IRV and pure FPTP is a 
deceptive one: that's not the real choice being 
made. The real choice has been between IRV and 
TTR, and FairVote has avoided making any 
performance comparisons except for misleading ones.