[EM] Why the concept of "sincere" votes in Range is flawed.

[Abd ul-Rahman Lomax]

At 01:11 PM 12/2/2008, Terry Bouricius wrote:
>Jonathan Lundell wrote regarding Approval voting strategy:
>
>"It's also obvious that if, for whatever reason, I vote for candidate X, I
>should vote for all the candidates that I prefer to X."
>
>I note that Jonathan said the voter "should," rather than "would," which
>is an important distinction...
>
>That is not the only strategy, nor in some elections the most likely
>strategy for many voters. If I prefer Y to X but love Z, yet also see Y as
>the most likely challenger to my favored Z, I might well vote for X (since
>I prefer X to W, whom I hate, and want W to come in dead last) assuming X
>has little chance of winning or hurting Z any way, but skip Y, even though
>I actually prefer Y to X, so as to not risk hurting Z, and also vote for
>Z. You might say this is poor strategy, or illogical, but I am reasonably
>certain it is the kind of strategy that would occur all the time.

This is Z>Y>X>W.

Terry is saying that he'd vote Z=X>Y=Z. Because he doesn't want to "hurt" Z.

Terry, it should be remembered, is a co-author with Richie of a paper 
on IRV, introduced IRV legislation as a member of the Vermonth 
legislature, and has remained a prominent advocate of IRV.

IRV advocates are a bit obsessed with Later No Harm. I think the 
average voter doesn't think about it at all. When it is presented to 
them as a really good reason to oppose Approval or Bucklin, they may 
think, "That sounds good!" But they haven't considered the 
implications. The implications of LNH are that the voter will not 
reveal the information necessary to find a good overall compromise 
until the voter's favorite has been executed, er, eliminated. Only 
then, okay, now that this is simply impossible, may Z rest in peace, 
I supported him until the bitter end, I'll tell you that I prefer Y 
over all others. Of course, by this time Y might also have been 
eliminated, whereas if I'd been a little more forthcoming, I might 
have elected Y.

Instead, I get X or maybe even Z.

We see here that Terry has a strong preference against W. Is W a 
possible winner? The vote for X makes no sense unless W could win. Y 
is the "most likely challenger" to Z. How likely? Is this a three-way 
election? Looks like it. Those elections are very unusual.

Let's give these some names. Z is Nader. Y is Gore. X is, say, a 
libertarian, say Badnarik, and W is, well, W.

Terry gave the example generically, but I know he was active in the 
Green Party, and I hope he'll forgive me if I make some assumptions, 
just to discuss this.

Real election? Historically, Gore v. Bush. Nader is greatly preferred 
by Terrill. The vote that doesn't seem to make sense here is the vote 
for X. And since that is moot, this is, for all practical purposes, a 
bullet vote for Nader, with a token of support tossed to Badnarik, 
which must be seen as an extra-electoral statement. Its purpose is 
not election. If X were a real possibility (a possible "challenger to 
Z"), Terrill would not vote for X. (Gore was not "a possible 
challenger to Nader." Nader simply wasn't on the map as a winner. 
Take Gore off the ballot, does Nader win? No, Bush wins.)

Yes. A reasonably common voting pattern, very much equivalent to 
voting for Nader alone in Florida. The only difference is the token, 
harmless -- and cast only because harmless and moot for election 
purposes -- vote for Badnarik. Perhaps Terry wants to signal the 
major parties that they should pay more attention to libertarian 
principles, which I'd consider commendable. (I'm a "progressive 
libertarian" or "libertarian progressive." Between Obama and Ron 
Paul, I'd hesitate a moment, then vote for Obama. Boy would I wish it 
were Range so I could toss Ron Paul some points. Between Clinton and 
Ron Paul, it would get tougher. Probably Clinton, dammit. I.e., I'd 
really like an election system that allowed more expression. I am 
*not* worried about "getting my favorite." I'm much more worried 
about the overall process, and social unity, finding consensus on a 
large scale. If we can do that, we can do better than we can do by 
simply electing the best candidate in one election.)

However, if Terry would seriously regret the vote for Badnarik, 
waking up Wednesday morning to find out that he wasn't as unique as 
he thought and so many other voters voted that way that Badnarick 
beat all the rest, I'd suggest, he shouldn't cast that vote. As it 
was, though, and in similar situations, the contrary of what he said, 
it is quite logical. And harmless. It was an "insincere" vote cast 
for reasons other than determining the winner.

Normally, if there are no outside considerations like this, a voter 
"should not" have voted for X but not for Y.

Approval does not require voters to vote a non-moot vote. IRV in some 
places does. Mandatory full ranking, in Australia. You *will* vote 
for all candidates but one, so we can put together our absolute 
majority and pretend that we have a really fine democracy. Coercion 
as a basis for democracy isn't brilliant.

It is not at all automatic that voters will cast a "sensible" vote. 
It's their prerogative to vote how they choose. They are responsible 
for the results, morally.

Now, Terry, there is available quite a good compromise, that allows 
you to make your stand for Nader and participate in the real 
election. *Technically,* it doesn't satisfy Later-No-Harm, but close. 
It's Bucklin. It worked. Contrary to FairVote propaganda, Brown v. 
Smallwood did not outlaw Duluth Bucklin because of Later-no-Harm 
violation, that was a single sentence, dicta. There are numerous 
statements in the decision, and the reaffirmation of it, showing that 
the problem was a single voter, with a single vote, facing a 
multiplicity of votes from other voters. The decision was against the 
entire concept of casting alternative votes. Every 
non-FairVote-affiliated lawyer asked for an opinion agreed that IRV 
violated the Minnesota Constitution if Brown v. Smallwood stands.

Bucklin not only satisfies the stated purposes of IRV -- all except 
for the highly questionable LNH criterion -- but is cheaper, saw 
substantial use in the U.S. before, was very popular (read Brown v. 
Smallwood! the Court noted that it was deciding against popular 
opinion and, as well, broad legal opinion as well). Bucklin addresses 
the common objection of people to IRV, the inability to indicate the 
favorite. Bucklin says, okay, vote for your favorite, fine. If the 
favorite wins by a majority, great. We are done.

But if there is majority failure, we take the lowest vote-getter in 
terms of favorites (first preference) out back and shoot him. No, 
that's IRV. Or might as well be. Getting the lowest first preference 
vote is an unremediable sin, even if this candidate is everyone's 
second choice and the result would be unanimity. Instead, we will 
accept a plurality for the winner, who is usually the plurality winner, period.

With Bucklin, nobody is eliminated. The second preference votes are 
added in, and if a candidate has a majority (determined on the 
original basis, i.e, a majority of ballots contain a vote for this 
candidate), done. Likewise to the third preference and, as you know, 
the voter could add any number of candidates to the third rank.

Like, "anybody but Bush," i.e., every candidate on the ballot. 
Bucklin allows anti-Plurality, in effect, should someone feel this appropriate.

Thanks for the example, I'd been thinking about such votes with 
Approval and hadn't taken the time to explore it in writing. The 
example shows Approval working as designed; Terry, what you seem to 
have overlooked is that the "illogical" vote was simply dicta, with 
no real effect.

One more possibility, though, should be explored. The election is 
different, i.e., it is three-way. The Green Party, against all odds, 
has clawed its way up and might actually win this election.

How do you vote? The same way, probably. It's up to you. Which is 
more important to you, that Nader beat Gore or that Bush be defeated 
by one of them? Your choice. That's the choice we want you to make. 
You could express your sentiments and take actions more in accord 
with them, probably, with Range. Approval is very simple, it's not a 
highly sophisticated method, though the improvement over Plurality is 
enormous, probably more than we get with IRV. At far lower cost.

Bucklin, though, does the good that IRV could do, without the bad. 
We'd be more likely to extract that Gore > Bush vote out of you with 
Bucklin than with pure Approval. Approval should only be compared 
with Plurality, to start. IRV might perform better than Approval in 
some elections, because of the Later No Harm issue that you raise. I 
think that most voters don't vote that way, in fact. A Nader 
supporter, seeing the move from Plurality to what I'm now calling 
Open Voting -- Approval -- would probably be grateful for the 
opportunity to add a vote that will count, and won't see a vote for 
Gore as "harming Nader." At least not much. Certainly not in terms of 
election outcome. Nor in vote count, if it's Open Voting.

IRV has this one good thing: it is *usually* true that adding a 
second rank vote does not reduce the chances of the favorite. But 
that's only because the method has artificially made it impossible 
for that favorite to be elected. You not revealing your second 
preference vote means that your vote can't help someone else until 
your candidate is eliminated. But the same is true for other voters: 
their second preferences won't be revealed until their candidate is 
eliminated. It is *quite* possible that in a close election like you 
posited, Nader would fail to be elected because of those other first 
preference votes. So, sure, you won't be responsible for your 
candidate's loss. Unless, of course, you are the one who chose that method!

Imagine if voters could pick the election method. Ballot is a 
preferential ballot, say three-rank. Voters can vote on a question: 
shall this election be canvassed using Bucklin or IRV? Wouldn't that 
be interesting? Or, even better. The voter decides how their vote 
will be counted. Let me describe that:

You have a check box: do not reveal my lower preference votes until 
my higher preference votes are only for candidates not eliminated.

The counting method starts out as Bucklin. Except ballots with the 
check mark are not considered as having any lower preference votes. 
When Bucklin counting is done, and no majority has been found, then 
IRV counting begins.  Later-no-Harm compliant, fully, for the voters 
who choose it.

Actually, it's interesting. There is a related possibility: IRV, but 
with equal ranking allowed. IRV-Open Voting. Fewer spoiled ballots, 
for starters. May be voted as Approval or IRV. May be voted as 
Bucklin! (Don't want to harm your favorite, only rank one in first preference!

But three ranks could be adequate even for large elections. More 
likely to find a true majority than IRV.

Down side: counting costs (quite like IRV, though perhaps a few more 
second rounds avoided). Bucklin, just add up the votes, in three 
stages, all of them precinct summable. The first hybrid I mentioned 
would avoid IRV counting in up to half of the IRV elections that now 
need sequential elimination.