FairVote is like a politician who tells people what they want to
hear. That's the art of spin. When it gets repugnant is when what's
being said is false. A post to the Approval Voting list, from which
I'm still banned from posting, referred to an article in the LA
times. It's worth noting that the author of this article has some
correct ideas, and he has merely been misinformed about the truth.
The truth isn't rocket science, but it is simply that there are
implications that often are overlooked by those not familiar with a field.
Opinion
Instant runoff voting
Such an electoral system saves time and money, and ensures a majority winner.
By Blair Bobier
December 10, 2008
http://www.latimes.com/news/opinion/sunday/commentary/la-oe-bobier10-2008dec10,0,6664124.story
Two examples from the seemingly never-ending 2008 election showcase
the system's flaws. More than a month after election day, we still
don't know who won Minnesota's Senate race. In Georgia's U.S. Senate
contest, it took two elections and tens of millions of dollars to
produce a winner. Both races could have been resolved quickly and
with less expense using instant runoff voting.
"Less expense" is a joke. IRV is expensive, and in a very close
election could be extraordinarily expensive. I haven't studied the
particular election, but what happened here was that the election was
close, and there are always a pile of ambiguous ballots, and
resolving the ambiguities, the way that we count votes here, can be
tedious. There are better ways to count, see various proposals I've
made for Public Ballot Imaging. This would quickly reduce ambiguous
ballots to a specific set, and the range of effects that they could
have on the outcome could be rapidly determined. If it is close
enough, then, there would be a clear and open basis for a legal
challenge to whatever conclusion the election officials issue, and a
rapid means of resolving the issue.
IRV results in higher numbers of spoiled ballots, some of which would
be ambiguous, and there are many more opportunities for ties.
Elimination sequence can affect the next stage of counting. In a
strong two-party system, IRV will *usually* work, and the only likely
tie is among the frontrunners, but it is quite unclear that IRV would
have created a big lead for one candidate or the other. It depends on
the exact configuration of candidates, and how many voters don't
fully rank, and so forth. On average, vote transfers don't change the
first preference order, so in a very close election, it will often
remain very close after transfers.
Georgia apparently requires a majority, which is a huge safeguard
against certain common election failures. Again, I haven't studied
it, but I presume Georgia was top-two runoff. While special elections
cost money, it's the price of democracy, in fact. IRV, we can now
tell, produces less democratic results than Top Two Runoff, for
reasons I won't address here. But it's certainly different, probably
differs from "instant runoff" in about one out of three runoff elections.
With instant runoff voting, voters indicate their first, second and
third choices among candidates on the ballot. If a candidate wins a
majority of first-choice rankings, that candidate is elected. If no
candidate receives an initial majority of first-choice rankings, the
candidate with the fewest first-choice rankings is eliminated and
that candidate's supporters have their votes count for their second
choice. The process repeats until a candidate emerges with majority support.
Now, the bait and switch is set up. "Majority support." IRV
supporters repeat that phrase like a mantra, and it is directly
misleading, to the extent that we could say, with the ordinary
meaning of words, and the author does explore this, and he thinks
he's using the ordinary meaning from the argument he makes, it is
just plain false. "Majority" in elections means that more than half
the legal ballots, containing a valid vote, have voted for the
winner. Preferential voting, of which IRV is an example, though a
particularly poor one, can allow alternative votes which can be put
together, but "majority" still means the same thing. These methods,
in general, can discover a majority that would be missing if voters
vote sincerely in Plurality Voting, for their favorite alone. So, if
we insist on a majority, preferential voting reduces the need for
runoffs but does not eliminate it. In fact, until this year, which I
haven't examined yet, the large majority of "instant runoffs" held
did not find a majority.
Replacing the Georgia top-two runoff method with IRV would be a very,
very bad idea, a step backward, actually reversing older reforms,
moving away from democracy. Replacing Plurality with IRV, better, but
expensive, and there are other forms of voting, including
preferential voting, that are far cheaper and which perform better at
discovering majorities and at finding the best winner by reasonable norms.
The Georgia runoff was triggered because a Libertarian candidate won
3% of the vote and the Republican finished the first round a handful
of votes shy of a majority. In Minnesota, 16% of the votes went to a
third-party candidate. In both cases, had voters been able to
indicate their second choice on the ballot, we would have known the
outcomes of the races on election night, saving a second election, a
recount and lots of time and money.
He's absolutely dreaming. Has he looked at, say, the San Francisco
Department of Elections web site? They only issue first preference
counts immediately, because those are easy to count. Lots of IRV
elections, the ones that don't find a majority in the first round,
are still counting three weeks later. It's been said that San
Francisco didn't call the method "Instant Runoff Voting" because it
is far from instant. If those minor parties are equally arrayed on
the left and right, which is common, the vote transfers tend to be
roughly equal. It takes a strict three-party situation for the
situation to be different.
Instant runoff voting is also an important innovation because it
produces a winner who has demonstrated support from a majority of
voters. When a candidate wins election with less than majority
support, it means that a majority of voters have actually rejected
that candidate. That's not fair to the voters, and it undercuts the
legitimacy of the electoral process. It is also, unfortunately, a
common occurrence in California and national politics. Three of the
last five presidential elections, and three of the last four
gubernatorial elections in California, were won by a candidate who
failed to win the support of a majority of voters.
He is utterly unaware, it seems, that he's just given the best
argument against using IRV as a method to finally determine
elections. In most of the IRV elections held in the U.S. since the
current rash began, the election results show that less than a
majority of voters voted for the winner. We really don't know, in
fact, because many, even most, of the various votes cast in IRV are
not counted. (San Francisco does publish so-called "ballot images,"
which aren't actual images, they are representations of what the
op-scan machines concluded was the effective vote; various vote
patterns considered moot or illegal have been unexpressed. These
images do include all the votes, but they are certainly not all being
officially counted, and most of them have no effect on the result.)
In one San Francisco election, the winner had less than 40% of the vote.
Instant runoff voting is politically neutral. It might have resulted
in the election of two GOP senators in 2008 or a Democratic
president in 2000. Who would have won the Minnesota Senate race
using it is anybody's guess, but a winner -- regardless of party
affiliation -- already would have emerged, the preference of the
voters would be clear, and the winner would have a legitimate
mandate to govern.
Under circumstances which are relatively unusual, but which might
occur, based on my studies, in up to ten percent of elections,
particularly nonpartisan ones, IRV can elect a seriously poor
candidate, because the best, who might beat the IRV winner by 2:1 in
a real runoff, is eliminated because of being third place in first
preference votes. This happens easily when this "compromise winner"
is a centrist candidate, the best compromise. So he's got it
precisely backwards. Plurality has some obvious problems, but IRV
doesn't necessarily solve them and, by encouraging sincere voting (a
good thing in itself, but not necessarily a good idea with IRV, just
as it is not necessarily a good idea with Plurality), it fixes *some*
of the problems but makes others worse. IRV advocates generally point
to partisan elections for examples of the great job that IRV will
supposedly do, but IRV is only used, around the world, in strong
two-party environments, where it looks best. When you get a third
party that actually rises up in prominence to the point where it
could win elections, IRV results become highly erratic and can
produce a worse problem than is likely with Plurality when most
voters make the necessary compromises when voting.
Instant runoff voting is used by cities in Maryland, Vermont and
North Carolina and approved for use in Tennessee and Minnesota, and
it has been used for years in Ireland and Australia. With momentum
growing for a national popular vote to replace the electoral
college, the day may come when it is used to elect the president.
We, the people, deserve no less: a simple and civilized way to
ensure that the outcomes of our elections reflect the intentions of
our citizens.
He has no idea of the nightmare that could result with IRV being used
for a national election? I can think of only one election in the
world where it's used for that: the largely ceremonial office of the
Presidency. IRV works fine when a majority is found in the first
round. It works okay when only a single round of transfers are
needed. But it is probably the most complicated voting system to
count. It *sounds* easy. It is not. In other -- better -- voting
methods, you can count the ballots locally and transfer the sums for
central tabulation. Doing that with IRV requires that individual
ballot data be transferred, because you cannot just sum up the votes
in each rank, IRV doesn't work that way. Rather, the effective vote
in the second rank on the ballot can only be determined when the
first round results have been found; second rank votes are only
counted from the ballots containing votes for eliminated candidates.
Sometimes, when there are many candidates, there can be many, many
rounds of elimination, and an error in any one of them can ripple
through all the subsequent rounds. Cary, NC, ran a trial IRV
election, and the election officials seem to have considered it a
nightmare. Election security experts appear to agree that IRV is much
harder to audit, you cannot extrapolate results from samples.
For a simple example of a better method, nearly a century ago there
was a method used in a number of different places, and the political
science literature considered it an advanced method. It was called
Bucklin voting, or the Grand Junction system, after the Colorado town
where the inventor (Bucklin) lived and was politically successful.
However, the simple name, used in some places, was preferential
voting. This is the same name that IRV was known by, and is known by
in Australia. The ballot, as used in, say, Duluth, Minnesota, looked
the same as a three-rank RCV ballot as is used in San Francisco. But
all the votes are counted, if needed. If there is a majority of first
rank votes, done. If not, the second rank votes are added in. If a
majority, done. If not, the third rank votes are added in, and the
candidate with the most votes won. The vote *totals* can add up to
more than the number of voters, but, in the end, these are really
alternative votes, because only one of them is effective to create a
winner, all the others could be stricken and would be of no effect.
Bucklin, like IRV before in the U.S., was rejected, but not because
it did not work. There were various reasons; in Minnesota, in
particular, the Minnesota Supreme Court, in Brown v. Smallwood,
decided, against what was clearly the prevailing legal and popular
opinion of the time, that allowing voters to cast any kind of
alternative or additional vote was contrary to the Minnesota
Constitution. That decision was not emulated elsewhere, but any kind
of election reform faces tough enemies, those who don't like the
possible changes that will result. In the case of IRV, though, most
people seem to get it backwards. IRV makes the world safe for the top
two parties. No longer can minor parties spoil elections, which
*reduces* the power of third parties. Regardless of theory, this is
what is seen in Australia. IRV activists will point to Burlington,
Vermont, where a member of the Progressive Party of Vermont won the
mayoral race using IRV; but, in fact, this candidate was the first
round leader by a good margin, was very popular (had served in the
Vermont legislature), and would have won the election with any
method. I've been told that in that town, the top two parties are the
Progressive and Democratic parties. The election results don't
necessarily confirm that, the Republican did get a substantial vote,
but city elections tend towards the nonpartisan in any case, even
when party labels are allowed. People vote for the person more than
for the party.
The most immediate danger is that the only method which actually, in
some environments, guarantees a majority result, or could, is being
replaced with IRV, which doesn't do that. Top Two Runoff can be
improved. Using a better method for the first round in TTR would be a
great improvement. Bucklin, for example, is more efficient than IRV
in finding a vote for the winner from a majority of ballots, because
it will count all the votes. With IRV, a vote for the winner may be
concealed underneath a vote for the runner-up, so IRV never counts
that vote. If we want legitimacy, we should Count All the Votes. I'd
estimate that Bucklin would eliminate the necessity for a runoff in
over half of the present levels of runoff with TTR. IRV doesn't do as
well, except, of course, when we simply eliminate the majority requirement.
That the Ranked Choice Voting proposition in California eliminated
the majority requirement -- literally struck it from the code --
slipped by opponents. The voter information pamphlet claimed that the
winner would still be required to get a majority of votes. Yes, if we
exclude from consideration all ballots of those who did not vote for
one of the top two. In other words, if you voted sincerely, but not
for one of two candidates, even if you detested them both, you didn't
count. Only people who voted for the top two count.
Real runoffs are not like this. It's a different set of voters, and
voters have had an opportunity to consider the two candidates left on
the ballot. In some places, if the "wrong" candidate has been
eliminated in the primary, which can happen with top two runoff just
as it can with IRV, voters can still write in a vote in the runoff,
and write-in candidates sometimes win. A single runoff does not
guarantee a majority result, either, but a good voting method used
for the runoff, such as Bucklin, would make a failure to find a majority rare.
There are far cheaper and better ways to fix what is broken with our
elections. I'd recommend Gaming the Vote, by William Poundstone, to
any who wants to know more about this. There has been a tremendous
amount of misinformation spread about IRV in the U.S., by people who
haven't taken the time to really study the rather complex field of
voting systems. But there are some simple voting systems, that are
not hard to understand, that are known to work well, we do not have
to use the complicated, quirky, and expensive to count "Instant
Runoff" Voting. The very name was invented to promote the method
here, and was promoted by activists who settled, for political
expediency, on IRV as the method to promote, knowing, in fact, that
they were creating a much more complicated voting system than is
needed for single winner elections. Why did they do this? Their goal
is proportional representation, and IRV is single-winner STV, Single
Transferable Vote, which is actually a decent method for creating
roughly proportional representation. They realized that an obstacle
to the success of proportional representation here was the complexity
of the voting system, so ... if they could sell IRV as a substitute
for "expensive" runoff elections, they'd have a foot in the door, and
the next step would not be so expensive. It made some sense, and the
goal of proportional representation is good, but using STV for
single-winner elections has been known to be problematic since the
nineteenth century. And there are also better ways to do proportional
representation, though STV is the best method in actual use.
IRV is, in fact, a nineteenth century system. It's been around a long
time. It has also been used in the U.S.; one long-lived usage was for
political primary elections. It was replaced with Top Two Runoff!
Now, why was IRV replaced with Top Two Runoff, if the latter was not
better? I think that's a good question. I think that people who
recognize that we need election reform should be asking questions
like this, and doing some serious research to find answers, not just
grab the first idea that comes along. It is very, very expensive to
convert to IRV, and it can continue to be expensive to run the
elections. The continued expense is probably cheaper than real
runoffs, but real runoffs have huge advantages, in terms of basic
democratic values. Political parties wanted to get better results!
There are other reforms, considered by voting systems experts to be
better, one of which is actually free. Just Count All the Votes! I
call it Open Voting, a name which has yet to catch on, it is
generally called Approval Voting, but it really is just Plurality
with a tweak: strike the no-overvoting rule, the rule that doesn't
count ballots where, single-winner, the voter votes for more than one
candidate. Approval Voting is considered a very respectable method,
some experts think it the best (I don't agree, but it's close!). All
voting equipment and procedures can already count it. Most voters
will not add additional votes: this has been called a defect, but it
isn't. Additional votes are only needed by voters whose favorite is a
minor party candidate, or in certain other unusual situations
(basically three-party situations where voting the sincere preference
is considered to risk the election of the worst candidate from the
voter's point of view.) Most people can continue to vote as they
have, and those who support minor candidates can pretty easily
understand how to vote.
But Bucklin fixes the problem with Open Voting, that there is no
means to vote to show which candidate is the favorite. Hence Bucklin,
which is a kind of "instant runoff open voting," may be politically
more feasible. It actually has a slightly lower counting cost than
Open Voting, on average. But a more complicated ballot, with probably
three ranks. It can handle *many* candidates with only three ranks,
because voting for more than one candidate in a rank may be possible.
In Duluth, the third rank was Open. The first two were vote-for-one.
Summary: Blair Bobier, learn something about voting systems and about
how IRV actually works!
----
Election-Methods mailing list - see http://electorama.com/em for list info
