I responded later in this thread, I've now gone back to the beginning
of it. I have had, of course, a special interest in Brown v.
Smallwood, the reasoning (or lack of same) behind it, and the
implications. Contrary to a couple of years of propaganda on the
topic, Brown v. Smallwood clearly prohibited any kind of "alternative
vote," not just the specific form known as Bucklin voting.
Brown v. Smallwood did not follow the precedents of other states, nor
did it become a precedent for other states. That's one side of this
issue. I'd consider BvS ripe for reversal.
But the problem is that FairVote and its proxies have argued that BvS
is valid, but that it does not apply to sequential-elimination
preferential voting, because only one vote is considered at any given
time. Thus they have attempted, in effect, to use BvS to allow IRV,
but not the probably superior Buckling voting, Approval, or Condorcet methods.
In particular, while Bucklin, like Approval, does allow votes for
more than one candidate to be active at any given moment, as must any
Condorcet method (with an exception of sequential-ballot forms with a
true majority requirement, used in parliamentary procedure but never
for public elections), in the end, only one vote *at most* is
effective. All other votes are moot and could be struck from the
ballots without affecting the outcome. Thus these methods do satisfy
the basic requirement of one-person, one-vote, *in terms of
determining the outcome.*
FairVote has attempted to make a virtue of the very serious Center
Squeeze problem of Instant Runoff Voting. They will say, over and
over, "voting for a lower preference candidate can never harm your
first preference." (It's not always true, if there is a true majority
requirement for election, but that's an unusual detail. IRV usually
*replaces* majority requirements, being sold as a supposedly cheaper
substitute for runoff elections. But it was true for the IRV proposed
by Mr. Bouricius in Vermont, while he was still in the legislature
there. Even though the "promise" was, by the legislation, placed in
the ballot instructions.)
However, when Woodall proposed Later-no-harm as a criterion, he noted
that a referee had commented, expressing disgust with this criterion.
It does, in fact, strike at the heart of democratic process, which
ideally involves seeking broadly-acceptable compromises (with the
minimum standard of acceptance being a majority approval of an
outcome); in order to collapse this into a single ballot, reasonably,
it's necessary for lower preferences than first preference to be
considered. Hence STV, used for single-winner elections, eliminates a
very important aspect of electoral democracy; this is why Robert's
Rules treats IRV as they do (with some distaste), and they do *not*
advise allowing election by plurality, ever. I.e., the "IRV" that
they describe does not avoid all runoffs, just, possibly, some. A
true majority continues to be an election requirement.
STV used for multiwinner elections still has this flaw in the last
candidate elected, but it does a generally good job with all the
other candidates; thus STV used for PR, particularly with more than
just two or three seats elected in a block, is quite a respectable
method. It's just the single winner variant that is truly
problematic. The problem arises when candidates start to be
eliminated before all the possible votes for them have been
considered. And this elimination is necessary to satisfy
later-no-harm, a very recently invented election criterion that seems
desirable only to the shallow in understanding.
Consider a ballot as a means whereby neighbors negotiate some common
decision. If I refuse to allow my lower preferences to be considered
until it has become clear that my favorite outcome has lost, and if
this is enforced by the ballot analysis method, it is impossible to
find compromise candidates, and the decision may fail to be the most
broadly acceptable, even if my first preference is only slightly
better for me than my second. When electing representatives, this
seeking of compromise is crucial, for government and the people to
remain closely connected. Clearly, proportional representation
methods, if properly designed, are far more desirable than
winner-take-all, single representative per district elections;
however, short of that reform, single-winner elections that go to the
first preference plurality can be expected to tend toward partisan
loyalty and two-party rule, as per Duverger.
Now, if IRV did, in fact, find majorities, it would be an improvement
over Plurality. By applying the study of single-transferable-vote,
done with full preference listings, to IRV, the promoters of IRV have
made it seem that this method would be such an improvement. But they
have glossed over a serious problem: IRV, in actual practice, as
implemented in the United States, where a majority does not appear in
the first round votes, only rarely finds one after vote transfers. It
is thus *normal* that most voters have voted *against* the winner,
i.e., they voted for someone else. What happened?
Their votes were exhausted, discarded, as not containing votes for
one of the top two candidates. If you've been paying attention, this
should start to sound familiar. That's what Plurality does. In
Australia, it seems that they get around this by requiring all voters
to rank all candidates; however, this only produces an illusion of
majority support, by requiring voters, essentially, to vote for one
of the top two, or their vote will be discarded as informal. (I.e.,
the voters are required to vote for all candidates but one.) Where,
in Australia, voters may truncate (not rank all candidates) (Optional
Preferential Voting, as in Queensland), the law was changed to
eliminate the absolute majority requirement used elsewhere; instead
something like "a majority of ballots containing votes for continuing
candidates" is substituted. And it's well-known that the first round
leader almost always wins the election.
Bottom line: IRV frequently elects by plurality, just like Plurality.
In non-partisan elections, exceptions are rare. There is no example
in recent U.S. history of the usage of IRV (since the adoption in San
Francisco) where the first round leader did not go on to win the
election. On the other hand, in Top Two Runoff elections, there is a
"comeback" election about one-third of the time. The strong
implication is that in about one IRV election out of ten, a candidate
is elected who would have lost a runoff against one of the other
candidates. The very existence of this phenomenon, and its
implications, except for my own so-far little-known work, have been
generally ignored.
Top Two Runoff has an obvious problem, if the first round is simple
vote-for-one. Sometimes a compromise candidate fails to make it into
the runoff. This is really the same problem as IRV, but the problem
doesn't exist -- or is ameliorated -- under some election rules. In
particular, Robert's Rules, for runoff elections, does not allow
ballot restriction. Partly for practical reasons, and partly for
political reasons (favoring major parties), popular political
elections do; however, under default rules in some places, including
California, runoff elections allow for write-in votes. Thus no
candidate is ever actually eliminated. Further, because
runoff-elections often involve more highly motivated voters, the
possibility of a write-in winner in a runoff becomes more real than
we would ordinarily consider it to be. A recent mayor of Long Beach,
California, won re-election even though term limits prevented her
from appearing on the ballot. (Long Beach is not a small town,
population is 466,718). If a candidate is a Condorcet winner as
determined by sincere preference rankings, and if the preference
strength is significant, as it apparently was in this Long Beach
election, the Condorcet winner should prevail, even though
"eliminated" in the first round.
(For years, election methods theorists totally disregarded preference
strength, though it is crucial in understanding election method
performance. A criterion "failure" that is based on minor preference
is less serious than one, otherwise similar, where the offended
preferences were strong. When a Condorcet winner fails to make it
into a runoff, but a candidate who is only slightly less preferred
does, voters or their leaders would not be exercised to conduct a
write-in campaign. But if the preference is strong, they can and
would. And have. Likewise, if a voter has only a small preference
between two candidates in a runoff (perhaps likes both, or detests
both), the voter is not exercised to turn out and vote, thus exerting
a Range-like effect on runoff vote totals.)
But, of course, IRV eliminates the runoff, thus eliminating a crucial
democratic principle, considered essential, that no decision be made
without majority approval of those voting in the election. Now, with
the Long Beach election, no majority was found even in the runoff,
because there was an independent write-in campaign. (The runoff had
only the name of the runner-up from the primary election on it.)
That's a compromise. But TTR is far closer to the democratic
majority-rule ideal than any method confined to a single ballot could
possibly be. To make it perfect, one would have to continue to
require a majority, thus risking a lengthy series of ballots, though,
in actual practice that doesn't happen much.
(Obviously, Plurality similarly discards this principle. One might
note that deliberative bodies, even large ones like the U.S. House,
would never use Plurality to elect their leaders. They use the
*method*, i.e. the ballot is vote-for-one, but with that crucial
detail: if there is no majority, the election fails and must be
repeated. Thus *the method is Condorcet-compliant, if there is a
Condorcet winner.* And TTR, a method which is not uncommon in the
U.S., likewise meets, *substantially,* the Condorcet Criterion if
write-ins continue to be allowed. IRV is replacing that method with
something inferior.
This is pernicious. But Brown v. Smallwood was a bad decision and
should be overturned, fully. The danger is that FairVote arguments
would result in a warped application of Brown v. Smallwood, to allow
IRV but not superior preferential or other advanced methods. Further,
wherever IRV has replaced top-two runoff, it should be made clear and
publicized that this is giving an inferior result about one-tenth of the time.
I only did a preliminary study, but Bucklin voting would find more
majorities than IRV does, because it does end up counting all the
votes, unlike IRV. (That IRV only counts some of the votes and not
others is an equal protection argument that could be advanced against
it. In Bucklin, all votes are treated equally. Bucklin is a kind of
sequential Approval voting.) My estimate is that about one-third of
the "majority failures" of IRV would be handled by Bucklin. I don't
have proof of it, but there is a good chance that most of the
possible "comeback elections" would have an unresolved majority
failure in the first round, under Bucklin. Bucklin, because it does
not eliminate candidates, but continues to count all votes to the
end, would, I believe, do a better job of finding a set of two
candidates to present in a runoff than would IRV. It is highly
unlikely that a possible runoff winner wouldn't be one of the top two
in a first round.
I will examine the affidavits posted by Kathy Dopp, separately.
----
Election-Methods mailing list - see http://electorama.com/em for list info
