[EM] Burlington Vermont repeals IRV 52% to 48%

[Abd ul-Rahman Lomax]

At 10:00 PM 3/5/2010, Dave Ketchum wrote:
>On Mar 5, 2010, at 8:34 PM, robert bristow-johnson wrote:
>but i *do* think that Approval *does* suggest the possibility of
>>bullet voting.  even though we're limited to 6 in my county, the
>>weird way that the Vermont State Senate is elected is that all state
>>senators are elected at large in your county.  so then more populous
>>counties have more state senators than smaller ones (they all get at
>>least 1).  anyway, we vote for up to six out of a zillion candidates
>>since each party proffers 6 candidates, we have Progs and Greens,
>>and there are independents.  the six candidates with the most votes
>>are elected.  what if there is one, maybe two candidates that you
>>really think should be elected?  i almost never vote for all 6.
>>usually just 2.  but it's a strategic vote.  and since i didn't hit
>>the limit, it's practically no different than Approval voting.  i
>>cannot see how Warren and company claim that it's less strategic
>>than Condorcet.

Approval voting is a single-winner system. That you don't cast your 
six votes in plurality at large, with six winners, is a demonstration 
of the flaws of plurality at larte (and of STV as well), it is not 
approval voting, which allows more votes than winners. There is a 
form of approval voting, proportional approval voting, and other 
forms, that are approval. You can cast just one vote and, provided 
you don't waste your vote for a loser (!), your vote will be just as 
effective as multiple votes, but multiple votes, assuming you vote 
for them all, allows alternate effectiveness. I.e., it's harder for 
your vote to be wasted.

>Approval's excuse for living is to allow voting for more than one,
>without encouraging or demanding doing such.

Right. Approval is not a perfect voting system (though in "repeated 
ballot" without eliminations, seeking a majority, it may be the best, 
with Range -- approval with fractional votes allowed -- being useful 
as a variation that might be a bit better.

As to strategic voting, it should be kept in mind that the modern 
"invention" of Approval voting by Steve Brams was as a 
"strategy-free" system. The term strategy there was a technical term. 
Of course there is strategic involved in Approval Voting, as with 
about any voting system. I.e., an ideal way to vote, that may not be 
the same as some supposed knee-jerk "sincere vote."

The term strategic meant that the voter gains an advantage by 
reversing preference. The term was developed for use with ranked 
voting. Both Approval and Plurality involve placing candidates into 
Approved and Disapproved sets, but Plurality only allows one 
candidate in the Approved set. Approval allows any number or even all 
but one, or even all. (If a majority is required, it is a reasonable 
vote under some circumstances, to vote for all candidates on the 
ballot. Or to vote for Anybody But Joe, by voting for all but Joe. Or 
to vote for None of the Above, by voting only for a write-in 
candidate, who, with Robert's Rules of Order, anyway, could be 
literally None of the Above. The vote counts, as all votes should. 
It's a vote against all the candidates.

Approval is quite clearly an advance over Plurality, and it allows 
"sincere voting" in that the vote is either purely sincere in the 
ordinary meaning, or, if the voter chooses the lesser of two evils 
for, say, two frontrunners, the vote is not *insincere*. It doesn't 
mean that the voter does not have a preference, it means that the 
voter has placed both candidates in the Approved set, and that 
decision is a strategic one (if the voter actually has a preference). 
Whether we make that decision depends on our perception of the 
electorate, the election probabilities.

There is a voting method which requires less such consideration, that 
would, in almost all circumstances, allow simpler sincere voting, and 
that's Bucklin, which might as well be called Instant Runoff Approval 
Voting, or, better, Instant Repeated Ballot Approval Voting. 
("Runoff" implies candidate elimination, which is what causes the 
problems with Single Transferable Vote, as Robert's Rules of Order points out)

If you have a sincere preference between two candidates, your 
favorite (not a frontrunner) and, say, a frontrunner, there is 
practically no reason in reality to not rank your favorite above the 
favored frontrunner, with Bucklin. In most Bucklin elections, all the 
ranks were collapsed, if I'm correct, so it reduced to pure Approval 
Voting. Later-No-Harm considerations means that you would not rank 
other candidates below your favorite if your favorite is a 
frontrunner (most people!) unless your preference strength was 
actually low. This is quite safe if a majority is required.

It's been overlooked, but it was also a perfectly acceptable 
strategy, if you were actually willing to accept another candidate, 
but had a significant preference for your favorite, to rank the other 
candidate in third rank, leaving the second rank blank. It's fairly 
clear that some voters, in real Bucklin elections, did, in fact, do 
that, as I recall. (Basically, more votes showing up in third rank 
than showed up in second, whereas truncation would produce -- and did 
normally produce -- the opposite effect.)

The practical reality of Bucklin was that it was a Range method, in 
effect, and that could be encouraged by using a Range Ballot for it. 
Very simple, with Range 4. The Bucklin ranks are ratings of 4, 3, 2, 
and 0. So add the missing rank of 1. It's a disapproved rank, so it 
won't be used to determine a winner, but it will be used for Range 
analysis, which might be useful under the following conditions:

Multiple majorities: whether to accept the vote leader or not might 
depend on Range results. (The alternative would *not* be the Range 
winner, it would be a runoff.)

Analysis of leading to possible victory declaration without a 
majority, based on clear evidence (including Range data) that if a 
runoff were held, the result would be the plurality leader.

It's also possible that Range data could participate in determining 
the identity of two or three candidates on a runoff ballot. If it's 
three, then there should be a preferential ballot there, as well. 
Maybe even if it is two, if write-in votes are allowed. (Very rare 
thtat this would make a difference, but, obviously, if write-in 
candidates ever win, or if they ever create a spoiler effect, *and 
they do, it's happened*, then preferential ballot is in order. 
Bucklin is much better than Approval, and if Bucklin allows equal 
ranking in first rank (as well as the lower ranks), it *is* approval 
voting if the voter prefers to vote that way.

Basic principle: give the voters power to choose how they want their 
vote to count. And then count all the votes. Even count votes that 
aren't used to determine the winner, it's rude to suggest and allow 
voters to vote and then not count their votes! And valuable 
information is lost.

Bucklin sets up a control device that casts their vote in a series of 
approval elections, where voters, according to the strength of their 
preferences, decide where to compromise. If the ballot is Range 4, 
that's four rounds that might be used. The last round would be 
counted, and used for various purposes other than direct 
determination of a winner in the first round, but in a runoff round, 
it would, in fact, be counted, perhaps, though it would be rare that 
it would make a difference. Basically, because a plurality result is 
going to be accepted, that lowest rating is the last chance for a 
voter who has not voted for either of the top two to cast a vote in 
that contest. And, with Bucklin, it's a full vote.

(But Okalahoma Bucklin made these fractional votes, which also makes 
sense, for Range analysis. Oklahoma was thus one of the few places to 
implement Range Voting! San Francisco, it seems, also tried this. 
Fractional votes! One of these days, some enterprising student of 
voting systems will research newspaper archives and find out what 
happened! The limited story we know is very strange, and quite 
different from what we'd have expected from other Bucklin 
implementations, but what if there ended up being only two candidates 
on the ballot! -- or only one of any substance!)