[EM] The truth about Instant Runoff Voting

[RLSuter at aol.com]

This is a reply to Dave Ketchum's suggested revisions of the
commentary on instant runoff voting (IRV) that I posted on
January 3. I want to thank Dave for his suggestions and
apologize for not replying sooner. I also want to thank
Forest Simmons for the suggestions he posted January 4.

Instead of responding point by point to those suggestions, I
just want to explain that when I wrote the commentary, I had in
mind a very general audience, most of whom would not have
nearly the mathematical sophistication of most subscribers to
this list, nor would many have the time or patience needed to
understand the details of either IRV or Condorcet voting (which
I called instant round robin voting/IRR) or the methods by
which votes are tabulated. Most general readers also wouldn't
have the patience to consider examples of election results
that would cause problems for IRV or IRR. So I didn't want to
include specific election result examples or use technical
terms like "matrices" or "pairwise" or "two dimensional
arrays." Doing so also would made the commentary too
long for most readers to want to take the time to read it.

But by being careful with language and explanations, the
basic concepts of IRR and IRV can be conveyed without any
technical language or election result examples. Matrices, after
all, are needed only as an efficient means for recording and
tabulating voting data. They are not needed to explain how the
winners of simulated one to one contests can be determined
from ranked ballots. That's very easy. If more voters rank
candidate A higher than candidate D, then A would defeat
D in a one to one contest. Almost anyone can understand
that.

In fact, by describing IRV as using ranked ballots to simulate
a series of runoff elections and by describing IRR as using
the same kinds of ballots to simulate contests between each
candidate and every other candidate, the two methods are
equally easy to explain, contrary to the claim of some IRV
advocates that IRV is easier to explain than IRR because
people are already familiar with runoff elections.

Dave did point out at least one thing that would have been
good to include in the commentary -- that "with IRR the ballot
can permit voters to have multiple candidates share the same
ranking." Still, when explaining the two voting methods to
people who aren't familiar with them, I think it helps to begin
by saying that they both use the same kinds of ranked ballots,
even though that's not quite correct. But it's close to correct,
and if there is time to explain the methods in more detail, it's
easy enough to add that IRV ballots and IRR ballots are
actually slightly different in that with IRR ballots, voters can
rank two or more candidates equally while with IRV ballots,
voters usually are not able to do that.

-Ralph Suter


In a message dated 1/18/05, Dave Ketchum wrote:

I left in the original text, to show the changes.  To read as revised,
delete the indicated number of original pars when you see "R 1", "R 2", etc.

On Mon, 3 Jan 2005 16:40:53 EST RLSuter at aol.com wrote:

 > This is something I posted on an anti-war list in response
 > to the post of an editorial advocating IRV. There is
 > nothing new here for most people on this list, but some
 > people may find it worthwhile to post to other lists.
 > I assume that little if anything I say here will provoke
 > controversy on this list, but if you think I've gotten
 > anything wrong, don't hesitate to say so.
 >
 >
 > -Ralph Suter
 >
 >
 > The truth about Instant Runoff Voting
 >
 > A recent newspaper editorial advocating Instant Runoff
 > Voting began as follows:
 >
 >    What if we were able to streamline our
 >    city and town elections, make them less
 >    expensive, and at the same time, ensure
 >    that the winner garnered a majority of
 >    more than 50 percent of the vote? [1]
 >
R 1
 > The truth is that Instant Runoff voting (widely known by
 > the abbreviation IRV) can't come close to ensuring this.
 > The only way a candidate can be said to have won a majority
 > of the vote is if it can be shown that she or he would
 > defeat every other candidate in a one to one contest.
 > Voting experts have a name for such a candidate: the
 > "Condorcet" winner, named after the 18th century French
 > mathematician, political philosopher, and activist (he
 > lost his life in the French revolution) who is now best
 > known for his advocacy of what is now widely called
 > Condorcet voting.
 >
 > The truth is that Instant Runoff voting (widely known by
 > the abbreviation IRV) can't come close to ensuring this.
In fact, if "the vote" means all the ballots cast (which many
of us would expect), stubborn voters can make it impossible
for any method to garner a majority (suppose about 1/3 vote
for each of A, B, and C - IRV will discard the weakest
candidate and the candidate with the majority of the
remaining 2/3 will win).
We will talk below of a method that uses the same ballot as IRV,
usually picks the same winner as IRV, and the rest of the time
does better at satisfying the above newspaper editorial - the
 > "Condorcet" winner, named after the 18th century French
 > mathematician, political philosopher, and activist (he
 > lost his life in the French revolution) who is now best
 > known for his advocacy of what is now widely called
 > Condorcet voting.
 >
 > A better name for Condorcet voting would be "instant round
 > robin voting" (call it IRR). The IRR method would use the
 > same kinds of ranked ballots as IRV and therefore would
 > require the same kinds of voting equipment. (Both would,
 > for all practical purposes, require computerized vote
 > tabulators, because tabulating either IRV or IRR elections
 > by hand would be prohibitively time consuming and prone
 > to human error.)

An advantage both methods share over many others is their
combination of voter power with simplicity for voters - simply indicate
as many candidates as you choose, in order, starting with most
liked.  A voter can indicate a single candidate, as with
Plurality, as well as more when desired.  With IRR the ballot
can permit voters to have multiple candidates share the same ranking.
 >
R 2
 > The one difference between the two methods -- and it is
 > a very big difference -- is in how the ranked ballots
 > are tabulated to determine the winner. The basic concepts
 > are very simple. With IRV, the ballots are tabulated to
 > simulate a series of runoff elections. As soon as a
 > candidate gets a majority in one of those simulated
 > runoffs, that candidate is declared the winner.
 >
 > With IRR, the ranked ballots are tabulated to simulate
 > contests between each candidate and every other candidate.
 > If one candidate defeats every other candidate in those
 > simulated contests, that candidate is declared the winner.
 > Such a Condorcet winner is the only candidate who can
 > ever be said to have truly "garnered a majority of more
 > than 50% of the vote."
 >
 > The one difference between the two methods -- and it is
 > a very big difference -- is in how the ranked ballots
 > are tabulated to determine the winner. The basic concepts
 > are very simple. With IRV,
the first-choice votes are counted, and if a candidate has the support of
the majority of the voters, the election is over. If no candidate has the
support of a majority, the candidate with the least votes is dropped (as
are any votes with no remaining candidates) and his supporters' votes are
distributed among their second-choice candidates. So it goes, until a
candidate has more than half the remaining votes.

Note that this IRV tabulating effort requires that each vote pattern be
accessible, though votes with identical patterns can be grouped.  Gets to
be a lot of data if a large district with many candidates.

In Condorcet (IRR) voting, each vote is recorded in a two-dimensional
array referred to as a pairwise matrix. If candidate A is ranked above
candidate B, then the element in the A row and B column gets a 1, while
the element in the B row and A column gets a 0. The pairwise matrices
for all the votes are summed for the entire district, or any part of
interest, whether precinct, state, or other.  The winner is determined
from the district sum.

Note that public elections often consist of two sets of votes - those done
on election day and absentee and others completed later.  After polls
close, sums can be totaled and made public.  This information gives
apparent winners and, with count of late votes expected, indicates chance
that absentee votes may change the winner.  Note also that these sums
indicate comparative strengths among all the candidates.
 >
R 2
 > The problem is that both IRV and IRR have well known
 > weaknesses. In fact, voting experts have long known that
 > there are no voting methods for choosing among three or
 > more candidates that have no weaknesses. Therefore, voting
 > method researchers try to find methods that have the
 > fewest or least objectionable weaknesses.
 >
 > The big problem with IRV is that the winner is very often a
 > candidate who would have lost to one or more of the other
 > candidates in a one to one contest. This is not just an
 > occasional problem but would likely happen frequently.

Both methods inspire complaints.  IRV can get in trouble by
making decisions without looking at complete votes - not clear
how often this might happen, but the complaints can be
expected to be loud and bitter when it does.  Suppose
the votes are:
   45 MA - these care only about their major party candidate.
   28 MB - these care only about their major party candidate.
   27 TP, MB - these have been sold by third party candidate
      TP, but keep their major party candidate for insurance.

IRV and IRR will see TP as having too few votes to play and
MB's 55 votes will win.

Now suppose two more voters heard and liked TP's campaigning,
so the vote count became 45/26/29:
      IRR will still see TP's count as low and award MB.
      IRV will discard MB's 26 as low and award MA - when the
voters have said nothing about liking MA better.
 >
R 1
 > The big problem with IRR is that sometimes there is no
 > candidate who will defeat each of the others. In such
 > cases, no matter which candidate is declared the winner,
 > there would be at least one other candidate who would
 > defeat the declared winner in a one to one contest.
 >
IRR can display in its matrices when its voters fit in
three or more groups that disagree as to best candidate and
are strong enough to approach a tie among their choices -
perhaps A>B AND B>C AND C>A.  The IRR implementation must
be prepared for a "reasonable" resolution of such cases,
plus an explanation that will hold the "experts" complaints
to a tolerable noise level.  The resolution is a solvable
challenge; finding a salable explanation that IRR is simply
responding to real life collections of voters seems to be
accepted by some as unsolvable.

R 2
 > No one can say for sure how often these problems would
 > occur in actual elections run with either IRV or IRR. My
 > guess is that a non-majority IRR winner would be much rarer
 > than a non-majority IRV winner. But advocates of IRV might
 > disagree, or they might say it doesn't matter because IRV
 > is superior to IRR for a variety of reasons. I won't try
 > to settle that disagreement here.
 >
 > Unfortunately, advocates of IRV, IRR, and some highly
 > regarded non-ranking methods that have the advantage of
 > being much easier to tabulate (most notably, Approval
 > Voting), almost never get to debate except among very
 > small groups of specialists. The biggest organizational
 > advocate of IRV, the Center for Voting and Democracy,
 > has essentially prohibited debates under its auspices.
 > Its leaders almost never engage in debates except when
 > necessary to respond to articles in prominent publications
 > such as Science News [2] and Scientific American [3]
 > in which different methods are advocated or favorably
 > discussed.
 >
Learn a bit more from articles in prominent publications
 > such as Science News [2] and Scientific American [3]
 > in which different methods are advocated or favorably
 > discussed.

 > In short, Instant Runoff Voting is not the panacea its
 > advocates have made it out to be. If IRV advocates are as
 > serious about democracy as they claim to be, they will
 > cease to be merely advocates of a particular voting method
 > but will also advocate and participate in honest, public,
 > democratically conducted debates between IRV advocates
 > and advocates of other voting methods.
 >
 > Ralph Suter
 >
 > (An attendee of the 1992 founding meeting
 > of the organization that later became the
 > Center for Voting and Democracy)
 >
 > [1]"Runoff vote system makes good sense", Southwest
 > Florida News Press, January 3, 2005, www.news-press.com/

 > apps/pbcs.dll/article?AID=/20050102/OPINION/501020401/1015
 >
 > [2] "Election Selection: Are we using the worst voting
 > procedure?", Science News, November 2, 2002, p. 280.
 > www.sciencenews.org/articles/20021102/bob8.asp
 > [mainly about Approval Voting]
 >
 > [3] "The Fairest Vote of All", Scientific American,
 > March 2004, www.sciam.com  [mainly about Condorcet
 > voting, which is called "True Majority Voting" in the
 > article]

-- 
   davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
   Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
             Do to no one what you would not want done to you.
                   If you want peace, work for justice.