[Election-Methods] Bullet Voting in the wider media

[Abd ul-Rahman Lomax]

At 07:04 PM 10/8/2007, Jonathan Lundell wrote:
>>Are we saying that a bullet vote for Abraham Lincoln is insincere?
>>Why? The voter has essentially set an approval cutoff between
>>Abraham Lincoln and Adolf Hitler. In this case, that isn't even
>>questionable, it is quite sincere.
>
>Which is exactly what I meant by "unfortunate shorthand" above. Per
>your premise, our voter as dictator would fill the seats which Mr
>Lincoln and Mr Khan (or is it Mr Genghis?), assuming that both seats
>had to be filled from the candidate list.

Actually, no. As dictator, he would fill one seat and leave the other 
vacant until he found a better candidate!

This is one of the most offensive practices in actual elections, the 
assumption that the office *must* be filled. Robert's Rules dislikes 
that any action be taken without the support of a majority, voting 
explicitly on the question. What is interesting about Approval is 
that the winner clearly has that, the majority has decided to support 
the winner, there isn't any doubt about that. True, because of 
imperfect knowledge, the majority might actually prefer another 
candidate, but ....

Well, I've seen it in action. A group had a majority preference 
(actually probably supermajority) for one thing, the status quo. A 
minority proposed that this be changed This was a group which valued 
group unity (and I'd claim that we should similarly value social 
unity overall), and so an Approval Poll was taken. Which if the 
following options would be acceptable to you? The majority could have 
bullet voted. But they did not. And while the status quo got 
something like a two-thirds vote, there was another option that got a 
*unanimous* vote, less only one. The obvious was then done, a motion 
was made to adopt the new option, and it passed *unanimously*.

So my view of the majority criterion is colored by that. I've come to 
think that majority consent to any decision is *crucial*, and this is 
fully in line with Robert's Rules. So if a candidate does not get a 
true majority, the election *fails*. Robert's Rules also dislikes 
top-two elections, which essentially *force* a majority. Rather, it 
wants "repeated balloting" until the group finally figures out what 
it collectively wants.

>  With STV, there's no
>problem expressing that preference. But with plurality or approval
>voting (or range, I suppose), the voter is forced to truncate his
>preferences to maximize his most-desired result: that Lincoln be
>elected, regardless of what happens to the second seat.

We have to see "approval" as the next color in a spectrum of methods. 
This is the hierarchy as I see it:

1A. Vote for one only (equivalent to Yes for one, No to all others.
1B. Vote on each candidate as a Yes/No, as if this were a question, 
Shall this candidate be elected?

Precedent is established in the second case that, if more than one 
candidate gets a majority, the one with the most Yes votes wins. 
(See, say, the Nevada State Constitution on the question of multiple 
conflicting ballot questions.)

 From here, where do we go. There are two options.

2. Allow fractional votes. This, of course, is Range.

Why did I consider 1A and 1B to be variants of the same method? Well, 
they are counted the same, just add up the votes. Further, if we have 
a show of hands vote in a face-to-face meeting, there is no 
prohibition against voting for more than one candidate. I'm *sure* it 
happens, I've never seen a rule against it, and, in fact, there is no 
reason at all to prohibit it. In Robert's Rules, when the clerk is 
instructed to discard overvotes, there is a reason given: because the 
intention of the voter cannot be discerned. That *assumes* that 
overvotes are prohibited. It certainly is not a reason to prohibit 
them! I have, in fact, never seen such a reason that made any sense; 
as the dissent in Brown v. Smallwood (Minnesota) noted, there was no 
violation of one-person, one-vote in Bucklin (which is "instant 
runoff approval," i.e., starts out with a single-vote rank, if no 
majority, next rank votes are added in, if no majority, third rank 
votes, which are not restricted to one vote, are added in.) It was 
only possible to discard overvotes with written ballots....

I said there were two options: the other option is using ranks, 
preference order. The big problem with preference order is that 
preference strength isn't involved; a huge preference is treated 
identically with an almost nonexistent one. And in real-world 
decision making on a small scale, this is blatantly a poor way to go. 
Does it get better as the scale increases? I don't see why it would! 
This is where most of the election paradoxes and difficulties arise. 
The big problem with Plurality -- and Approval is really a plurality 
method -- was the restriction to a bullet vote.

Bullet voting makes sense if you are a supporter of a frontrunner in 
a two-party system, that is, your favorite is one of the top two, and 
any third candidate isn't viable as a winner. But that doesn't mean 
that the option of additional votes would be useless; for the spoiler 
effect arises with only a small percentage of voters, typically those 
drawn to a third party. Allowing multiple votes makes a big 
difference for those few percent, and efficiently eliminates the 
spoiler effect. Without forcing a false majority, as 
ballot-discarding IRV methods can do. Robert's Rules variation of 
preference voting, as I read it, and consistently with other 
parliamentary rulings on the subject, does *not* discard ballots, as 
it eliminates candidates, if a ballot is exhausted, that ballot is 
set aside, but continues to be considered as determining the basis 
for majority. Unless the bylaws allow election by a plurality. If so, 
then the winner of the final round prevails. But not by a majority. 
By a plurality.

>If you don't want to use the term "sincere" here, that's fine by me;
>let's use something else. Let's find some term that describes an
>ideal method in which a voter can express his true (dictatorial,
>perhaps benevolently so, perhaps not) preferences without worrying
>that there's some way of voting otherwise to achieve a better result.

Well, there is such a method, actually. First of all, you've got to 
collect the necessary data, and the only ballot that does that is a 
Range ballot. But you can analyze a Range ballot as if it were a 
preference ballot with equal ranking allowed. There are two ways to 
go: with sufficient resolution, it can be a simple Range ballot, 
because a voter can maintain a preference of only one rating step, 
which is really pretty small if it is Range 100. It's still pretty 
small with Range 10! However, if the resolution is low, the device 
would be used of having a preference indicator that does not alter 
the Range vote. I.e., you could vote two candidates as perfect 10s 
but still prefer one.

But, it turns out, you would be unlikely to actually do that, in what 
I propose. Basically, the ballots are analyzed two ways: sum of 
votes, which determines a "Range nominee," and pairwise. If the Range 
winner is the Condorcet winner, and if the rules allow a victory by a 
plurality (I don't like that), then the election is over. There is no 
question about "plurality" if the Range winner is preferred by a majority.

But if the Range winner is beaten by another candidate, pairwise by 
preference, then there is a runoff. Now, some would say, ah, the 
pairwise winner would simply beat the Range winner. I don't think so, 
for two reasons:

(1) Some people like the idea of having a broadly acceptable winner.

(2) The runoff tests preference strength. When the Range winner is 
beaten in preference, there exists a special -- and relatively 
unusual -- condition. (Usually Range chooses the Condorcet winner). 
This is a minority with a strong preference and a majority with a 
weak one. The supporters of the Range winner must, under these 
conditions, have a strong preference, at least they voted that way. 
The supporters of the pairwise winner have a relatively weak 
preference. Therefore the latter have less motivation to trouble 
themselves to vote, they don't really care that much. The former, on 
the other hand, care very much and will be more strongly motivated to 
turn out. Compulsory voting ruins this particular model!

What this does is to test the sincerity of the Range ratings. If they 
were exaggerated, then the analysis above won't be true; the Range 
winner's supporters, with those exaggerated preferences, won't 
actually turn out as strongly as their votes would indicate.

But, it must be noted, such runoffs would probably be rare. And there 
are, in fact, ways to make runoffs much easier to handle.... we could 
have *much* better voting systems if we could cooperate to that end.

>>What bullet voting means, if deliberate, that the voter has such a
>>strong preference for the favored candidate winning that the voter
>>does not want to support any other candidate against him. While not
>>as drastic as the example I gave above, it merely indicates a
>>strong preference for the single candidate, strong enough that the
>>voter is willing to give up influencing a second seat. What's
>>insincere about that?
>>
>>There is a contradiction set up in every discussion I have seen of
>>the topic of strategic voting in Approval (and similar arguments
>>are made with Range):
>>
>>1. There is a voter who approves of two candidates
>>2. But only votes for one because the voter wants that one to beat
>>the other.
>>
>>Ahem. Those are two contradictory conditions! Part of the problem
>>is the use of the term "approval." I was just reading Voting
>>Matters and discover that I'm not the first person to suggest that
>>we are talking about voting, not approving. I might vote for
>>someone I rather heavily disapprove of, if I have no better
>>practical option. A Nader supporter might vote for Gore, even if he
>>thinks that Gore is just as much a tool as Bush, for there are
>>other issues, such as Supreme Court appointments, etc.
>>
>>My point is that a voter can set an approval cutoff anywhere the
>>voter pleases, and there is nothing insincere about it, in the
>>ordinary sense, nor, in fact, in the technical voting sense. What
>>has happened is that terms and measures developed for ranked
>>methods are being applied to cardinal methods. In a ranked method,
>>"insincere" has a clear meaning: preference reversal. That's easy
>>to define! But preference reversal never benefits the voter in
>>Approval, nor in Range.
>>
>>However, those who are actually advocating a ranked method, such as
>>"Instant Runoff Voting," can't stand the idea that Approval is not
>>"vulnerable" to insincere voting, so they must extend the
>>definition of "insincere" to include something else. Basically,
>>they posit an approval cutoff of their own, such that the voter
>>"approves of" two candidates, but only votes for one. And then they
>>call this an "insincere vote."
>>
>>Now, unless the voter is merely lazy, we have to say that the voter
>>voted for the candidate the voter preferred; that the voter placed
>>his approval cutoff between the two candidate utilities. There is
>>*nothing* insincere about this, and no way to truly apply the
>>concept of tactical voting to it. There is no preference reversal.
>
>There's a forced lack of expressiveness.

Compared to what? Not compared to Plurality!

>  It's exactly the "no better
>practical option" that you refer to above that I'm objecting to wrt
>approval and plurality voting. If the voter in your example were
>dictatorially picking the president, he'd choose Nader, and yet you
>have him voting for Gore.

We aren't dictators. We must find ways to compromise with others, to 
seek broad agreement, or we have badly divided societies, we can even 
have civil wars.

Look, I don't think Approval is optimal. It is cheap, it is easy, it 
solves the spoiler effect, and it simply remedies an old, habitual 
injustice. From there, we can move to better reforms. For now, 
immediately, where possible,

Count All the Votes!

>Again, if it bothers you to call that "insincere", I'm not going to
>argue over the term.

There is a whole philosophy of voting and the meaning of a vote. What 
is it? Do people have equal voting power? There is a problem with 
pure ranked voting: it equates strong preference with weak. 
Preference is actually a pairwise comparison, from a set of pairwise 
comparisons we construct a rank order.

What happens with alternative voting systems is that one full vote is 
exerted between, say, A and B, and then one full vote between B and 
C. There is, in fact, no gradation in this and it does not allow any 
clear expected measure of how satisfied people will be with the outcome.

In any case, there is a problem in the literature. "Strategic Voting" 
is being defined in a way that makes an infinitesimal preference 
equal to a strong one. To me, though, it's politically moot.

There is a simple reform that will not offend election officials. 
They already know how to count multiple-winner elections, just add up 
all the votes, and don't discard overvotes. We get Approval simply by 
stopping the practice of disregarding ballots that have overvotes on 
them. There are election methods and conditions where allowing voters 
more than one vote would be unfair; but that's not true with standard 
single-winner Approval. Then, want to set up ranks? I expect that 
what we will see is a range of experiments. We might see Bucklin 
again, the arguments of FairVote as to why it was dropped seem to be 
purely speculative projection, the most likely reason it was dropped 
was the same as why IRV was dropped in Ann Arbor: it was working, and 
election results changed, and the loser's party went after the method 
so they could once again rely upon vote-splitting. And we might see 
ranked ballots and Condorcet analysis. And we might see additional 
ratings beyond the 0 and 1 of Approval.

The Range 2 internet polls on MSNBC, of the presidential candidates 
in each party, were fascinating. They showed far more information 
about the candidates than standard plurality polls. You could see the 
spread .... and Ron Paul was way ahead when I looked in the 
Republicans. Computer nerds, fanatic Ron Paul devotees? That's an 
easy explanation, but I don't think that's it. I think the method 
allowed people to set aside questions of electibility and vote sincerely.

(Range N, in my notation, refers to a method with N preference steps. 
So Range 1 is Approval, Range 2, in the MSNBC case, allowed votes of 
-1, 0, 1. Except for how blanks were handled -- in this case the 
default vote was 0 -- this is equivalent to 0, 1, 2.)