WHY RCV IS UNSUITABLE AND AV IS NOT
By John Howard Wilhelm
Voter Choice Massachusetts and Rankmivote.com in Michigan, given
their concern to make our elections more open and representative,
advocate introducing Ranked-Choice Voting (RCV), aka Instant Runoff
Voting (IRV), into them. I have been studying the technical literature
on voting systems for over two decades during which I have interacted
with some of our best voting system specialists and find this troubling
for two reasons. First as a casual look at both organizations' websites
illustrates, both organizations clearly buy into the misleading and/ or
bogus claims often associated with FairVote's advocacy for RCV. And
second, I don't see how a factual, or evidence based, analysis of voting
systems validates the case for replacing our simple system of plurality
voting with its IRV plurality alternative.
If we want to move away from our increasingly dysfunctional political
system, changing our voting system is a necessary, though not sufficient,
condition for doing so. In thinking about alternative voting systems there
are two important considerations to take into account. First, does a voting
system have a strong tendency to elect the most representative candidate
given voters' preferences? And second, does it level the voting field for
third-party and independent candidates?
Under IRV voters are asked to rank candidates in order of their preferences.
These preferences are used to simulate a series of runoff plurality voting
elections in which each voter is permitted to have only one vote cast in
each round. If no candidate has a majority of the first vote count, the
candidate with the lowest number of first place votes is dropped and his
or hers supporters' second place choices are allocated in a second round
in which they are treated as equal to the other voters' first place choices.
Again if no candidate emerges with a majority count of votes, the process
is repeated until a winning candidate emerges with a majority of the vote
counts in a simulated plurality voting election.
The proponents of IRV, especially those associated with FairVote, assert
that "it will give all candidates a chance to compete and win." This assertion
is clearly the basis for Green Party members' support of IRV given that they
assume that this implies that IRV will open up our political processes to third
parties in our elections. But the assertion is misleading since the same is
true of our system of simple plurality elections as witnessed by the success
of a Jesse Ventura in Minnesota in living memory.
IS IRV REALLY THIRD PARTY FRIENDLY?
The issue here is whether IRV is more receptive to the success of third-
party candidates than our current system of simple plurality voting in
which the candidate with the most votes, not necessarily a majority, wins.
Both the empirical and theoretical evidence suggests that it is not.
Australia has used IRV to elect members of its House of Representatives
since 1918. But over a hundred years later, as data on the latest Australian
election clearly show, it has done nothing to move away from the two-party
duopoly of the Liberal/National coalition and the Australian Labor Party
or engender any meaningful participation of third-party members in that
legislative body.
Simulations of voting systems by Ka-Ping Yee found in a piece by William
Poundstone on his website titled "A Test Drive of Voting Methods" in part
suggest why. In his simulations based on an assumption of sincere voting,
IRV elections, with the exception of the monotonicity problem examined
below, give similar results to that of simple plurality voting.
But we know from the Gibbard-Satterthwaite theorem, that it is impossible
to devise a voting system in multi-candidate, single-winner elections that
can avoid an incentive to vote tactically or strategically. So clearly in the
Australian case of IRV elections. it is also a matter of how voters interact
with their voting system in a strategic sense. Until 1992 in Australia voters
in their national IRV elections were required to rank all candidates to have
a valid ballot. But after 1992 this was no longer.the case in the state of
Queensland. Data that Warren Smith presents on his website of "plumping,"
bullet voting in American parlance, in the 2009 Queensland federal election
suggest that a larger proportion of voters bullet vote for just one candidate
than is characteristic of approval voting elections. This may suggest that
in IRV elections Australian voters interface with an IRV election similar to
how voters respond to simple plurality multi-candidate elections.
Given its cultural similarities with the British it is interesting to compare
Australian IRV elections for its House of Representatives with British
elections to their House of Commons which are conducted under simple
plurality voting. In Australia to run for its House of Representatives, a
candidate needed to make a deposit of $1,000 Australian and beginning
in 2019 of $2,000 Australian, or $670 and $1340 US respectively, all of
which is given back if a candidate gets 4 % of the first place votes under
IRV or wins the election. In Britain all it takes to run for Parliament is the
signature of ten constituents and a 500 pound deposit, $659 US, all of
which is given back if a candidate gets 5% or more of the vote. That is,
ballot access for third party candidates is low in both countries.
But despite that, the electoral systems in both countries strongly favor
a two-party duopoly in their lower legislative bodies. In the 2016
Australian election to its 150 seat House of Representatives its two
major parties held 96%of the seats as they did in the 2019 election. In
Great Britain the dominance of its two major parties in its House of
Commons was 89.1% in its 2017 election and 87.3% in its 2019 election.
Based upon this data it would be hard to reject a hypothesis that simple
plurality voting is more favorable to third parties than IRV.
BOGUS CLAIMS AND DYSFUNCTIONAL BEHAVIOR
In my April 28, 2011 letter in the Financial Times just prior to the British
referendum on introducing IRV into their parliamentary elections, I pointed
out that claims as to the advantages of IRV turn out on critical examination
simply to be bogus. An example of this can be found in a piece following
the adoption of IRV in Maine elections by Rob Richie, the executive
director of FairVote, in the Winter/Spring 2017 issue of Green Horizon
Magazine. In the piece by Richie, it was stated that, "Starting in 2018
Mainers will be able to vote for the candidates they like the most without
helping elect the candidates they like the least."
That this is patently not true is illustrated by data on voters' rankings
of candidates from the 2009 Burlington, VT IRV election. In that election
it is clear from the data that by voting for their candidate the Republicans
helped elect their least favorite candidate, the Progressive winner. As
Andrew Jennings, a very knowledgeable voting system specialist, has
pointed out unless voters' first choice has a very strong chance of
winning or losing, voters cannot avoid the possibility of helping to elect
their least favorite candidate by placing their favorite first in their rankings
under IRV,
In many cases proponents of IRV seem to be unaware of the serious
pathologies that infest IRV and make it a very questionable voting
system for electing the most representative candidates. The following
link to an analytical, as opposed to a polemical, piece on IRV by Kathy Dopp,
does a good job of pointing this out: Realities Mar Instant Runoff Voting. In addition to that, the reader may
find my piece from the follow link http://www.nationalrenewal.org/node/43
on a comparison between IRV and approval voting a useful read. Both
pieces recognize that monotonicity failure is a serious pathology of IRV
Monotonicity failure can occur when a candidate who would have won can
wind up losing if he or she garners more first place support in the course of a
campaign. This would happen because the increased support the candidate
receives in a campaign can change the order in which candidates will be
dropped and votes reallocated in arriving at a result. In the technical literature,
this is referred to as violating the monotonicity condition--the idea that if a
candidate wins more support this should not adversely affect his or her prospects
For instance, suppose 21 voters are voting for three candidates: Alice, Bob
and Charlie. Eight voters rank the candidates Alice 1, Bob 2, Charlie 3; two
rank them Bob 1, Alice 2, Charlie 3; five rank them Bob 1, Charlie 2, Alice 3;
and six rank them Charlie 1, Alice 2, Bob 3. Since Charlie has the fewest
first-place votes, he is eliminated, and those six votes now have Alice in first
place, so she wins 14 to 7.
But suppose the vote were slightly different, and the two voters who put Bob
first had instead ranked Alice top (Alice 1, Bob 2, Charlie 3). Now Bob, with
only five first place votes, is eliminated and those five rankings then have
Charlie in first place, so Charlie wins 11 to 10. Moving Alice up in a few
rankings converts her from a winner to a loser, because in doing so there is
a change in which candidate is eliminated.
The upshot of monotonicity failure as Kathy Dopp's example illustrates very
well is that in its presence an IRV election responds to voters' preferences
in a perverse way. For voters it means that in casting a vote for their
favorite candidate or ranking their least favorite candidate last they cannot
know whether for sure that will help the one and disadvantage the other as
intended.
Work done by Professor Robert Norman and Joseph Ornstein at Dartmouth
College suggests that monotonicity failure can occur in a fifth of close three-
man IRV elections, precisely the type of elections in which IRV is touted as
a significant electoral improvement. To my knowledge the only solid empirical
evidence we have of this problem occurred in our 2009 public IRV elections.
As far as I can determine there were 9 such IRV elections in the US in that
year. Of the nine, at least two, in Burlington VT and Aspen, CO, elections
exhibited this problem. We don't know about the other 7 elections because
in most cases the data on voters' ranking of candidates is not available in
the published results. But it is important to point out, as Warren Smith has
done, that the more candidates in an IRV election, the greater the possibility
of such failures.
IRV does not solve the spoiler problem except in special cases like the
2000 Florida presidential election as the 2009 Burlington, VT.election
illustrates. In that election the Republican candidate, who could not have
won clearly played a spoiler role. Had he not entered the race the
Democrat candidate would have won over the Progressive by a margin
of twice the win of the Progressive in the final round against the Republican.
In situations where IRV does not take care of spoilers, candidates who
could not win but whose presence can change the outcome of an election,
the spoiler effect can be even more pernicious under IRV That is, shifts in
preferences among minor candidates can change the order in which
candidates are dropped in the runoff process and votes reallocated with
the potential of leading to dramatically different outcomes. This is the
source of the erratic behavior of RCV that Sir Michael Dummett identified
in his book "Principles of Electoral Reform," Oxford University Press 1997.
In addition to that, Kathy Dopp's point that IRV does not treat voters
equally is certainly an addition troubling one about IRV,
A SUMMARY AND CONCLUSION
In sum, with its dependence on the plurality voting system in which a voter
can only cast one vote at each stage of the runoff, IRV, despite the claims
of its supporters, does not get rid of the wasted vote, the spoiler role and the
consequences of vote splitting when voters vote sincerely as proponents
of the system maintain they can without worrying about the consequences.
Approval voting, in which voters are allowed to give one vote each to the
candidate or candidates they support with the candidate having the most
votes winning, does get rid of the wasted vote, the spoiler role and the
necessity of vote splitting among the more representative candidates. It
certainly is also third party friendly in that voters always have under the
system an incentive to vote for their favorite candidate, even if that
candidate is a third-party candidate. And it certainly does a better job
of reflecting voters' preferences in the outcome as I argued in my piece
cited above.
On these considerations, I do not see how one can argue that approval
voting is not a much better alternative to our system of plurality voting
than IRV, which may not even be an improvement over our existing system
of plurality voting.