[EM] To Donald on Approval & IRV

Sun Feb 10 17:41:53 PST 2002

I realize that it's better to not encourage Donald by replying to
him, but I reply once in a while:

Donald said:
\
The ABC methods are corrupt
methods. So, it follows that I would object to corrupt methods with the
same vigor that I object to corrupt rules in STV.

I reply:

Myerson, in an article referred to in Weber's article in the
Winter '95 _Journal of Economic Perspective_, studied the degree to
which some voting systems, including Approval, Plurality, & Borda
, discourage corrupt candidates.

Myerson found that Approval, in Myerson's model, drives corrupt
candidates out of the political system.

Myerson didn't comment on
how well IRV does, but to apply Myerson's test meaningfully to IRV,
some generalization of that test is desirable.

My Unanimously Unpreferred Candidate Criterion (UUCC) is a
generalization & simplification of Myerson's test for corruption
discouragement. Approval passes UUCC. IRV fails UUCC.

Adam: "If you please, give an example of an election where approval will
fail to produce a fair result."

Donald: Let's use the last Florida Presidential election. The numbers were
about: 44 Bush, 44 Gore, 8 Nader, and 4 Others. Now, what do you
consider to be fair results? Do you want a method that somehow, by hook or
by crook, will raise Nader's vote total up to be in the same range as that
of Bush or Gore?

I reply:

We don't choose a method based on who will win if we use it.
We choose on fairness issues such as voter freedom or not forcing
drastically insincere strategy.

Yes, Approval would increase Nader's vote total. That's bacause
many who like Nader best voted for Gore, in order to defeat Bush.
With Approval, everyone can fully vote for their favorite, and
no one has any incentive not to. (That isn't true of IRV).

Maybe it bothers you that making everyone strategically free to
fully vote for their favorite would increase Nader's vote total.
But try to judge the methods by objective standards and criteria
rather than whether or not you like the candidate that they'd help.

Donald continues:

If Approval Voting was the method of choice in the last Florida election,
Nader would have had a good chance of winning, provided at least fifty
percent of the Bush and Gore voters made the mistake of foolishly giving
Nader a vote too.

I reply:

Donald suggests, here & below, that in Approval, the Bush voters
would vote for Nader as well as Bush. :-)

Donald continues:

In order for the election of Nader to become more of a certainty, the Nader
voters should only vote for Nader, no other candidate. Let the Bush and
Gore voters be foolish (the Bush and Gore Voters are not going to be that
foolish).

I reply:

If the Bush & Gore voters are not going to be that foolish, then
voting for Nader only won't make Nader win. The statement that
you've just made contadicts your own claim.

You're right, they won't be that foolish. Foolish isn't even the
word for Bush voters who vote for Nader & Bush. Voters who prefer
Gore have no reason to vote for Nader in Approval. If one of the
2 extreme candidate has a majority of 1st choice preference, then
nothing will help Gore anyway. Otherwise, one extreme side needs
Gore, and they need him more than the Gore voters need that
extreme candidate. For instance, a victory for Bush would be worse
for a Nader preferrer than for a Gore preferrer.

It wouldn't make sense for Gore preferrers to vote for Nader unless
they consider Gore's & Nader's merit to be about equal, and consider
Bush to be absolutely unacceptable. But if they believe Bush is
absolutely unacceptable, then it would be difficult to explain why
they like someone like Gore, if they're that principled.

Donald: Approval Voting is a method that can be easily foiled, to the
extent that the method disappears. Take some example that is like your
example of three near equal candidates, say 1010 A, 1005 B, and 1000 C.

I reply:

Thank you Donald, for bringing up that example. Let's call the
candidates Favorite, Middle, & Worst, and let's say that all we
know is that their favoriteness totals are about equal. One thing
for sure then is that no one has a 1st choice majority.  Say, as
would be typical, that we have some idea that the Middle voters would
rank a particular one of the extremes in 2nd place, for the most
part. Lets say we expect them to rank Worst in 2nd place.

Now, say you're a Favorite voter. Should you vote Favorite in 1st
place? Why? He can't possibly win. Your best strategy is to vote
Middle in 1st place, and move your genuine favorite down to 2nd place.
It makes no practical sense to vote sincerely, with Favorite in 1st
place.

So your example is an example that shows how typical it will be
for voters to have strategic incentive to dump their favorite in IRV.

Donald continues:

In politics there is a rule that could be written; "If a ploy exists,
someone will find it and we can be sure it will be used to win an
election."

I reply:

There's no such thing as offensive strategy in Approval.
Nor is there any need for favorite-burying strategy, and that
distinguishes Approval from IRV.

Donald continues:

Now, the question is: "When using the Approval Voting method, what ploy
could be used by one faction to win the above election?" Answer: Your
faction should only vote for one candidate. (your candidate of course)

If the supporters of any one of these three candidates were to only vote
for their candidate, that candidate would have a very good chance of
winning.

I reply:

But you said yourself that the Gore & Bush voters wouldn't be
foolish enough to help the Nader voters in that way.

Sure, if you believe that your favorite has a win, then you have
good reason to vote only for him. Otherwise, it's to your advantage
to vote for someone else too, for the candidate for whom you'd
vote in Plurality, but also for everyone whom you like better.

Of course, the supporters of the other two candidates may do the same, then
the method of Approval Voting disappears and we are left with merely
Plurality.

I reply:

The IRVies always make that claim. Approval is very different from
Plurality because you can help Middle against Worst, if you feel you
need to, but, while doing so, you can also fully vote for Favorite too.

Unless  you consider your favorite a likely winner, it isn't good
strategy to vote only for him. Donald & other IRVies seem to believe
that voters are going become radically different kinds of voters
when we adopt Approval. The IRVie assumption is that Nader preferrers who 
now vote for Gore, and are willing to dump Nader for that purpose,
will suddenly lose interest in helping Gore if we adopt Approval, and
will vote only for Nader. That would be nice, but there's no reason
to expect it to happen. They'll vote for Gore, just as they do now,
but they'll also vote for Nader, as they should be able to. Approval
adds an important voter freedom to Plurality. Approval is Plurality
done right.

Donald continues:

The success of Approval Voting depends on deceiving most of the voters into
believing that it is safe to vote for more than one candidate.

If you're one of the millions of Nader preferrers who voted for Gore
in 2000, then you wouldn't believe that it's safe to not vote for
Gore, and you'd vote for more than 1 candidate.

Don continues:

The supporters of Approval are not being honest with the people when they
implies to them, "Vote for Bush or Gore if you must, but also give one of
your votes to Nader." Or, "Of course, you want to vote for Bush or Gore,
but you have plenty of votes, give one to Nader."

I reply:

I've never heard any Approval advocate make a voting suggestion
like that. Only Donald would come up with the suggestion to Bush
voters that they should vote for Nader in Approval.

Donald continues:

By replacing ranked preferences with actual votes, ...

I reply:

Actually we don't have ranked preferences yet, and so adopting
Approval wouldn't involve replacing ranked preference with unranked
votes.

Donald refers to the "ABC methods", to distinguish them from the
PIRV methods, consisting of Plurality & IRV.

ABC stands for Approval, Borda, Condorcet. And it's true that they
all have something in common: They don't just look at favoriteness.
When carrying out an important project, would you make irreversable
decisions based on only a fraction of the available information?
That's what the PIRV methods do.

A candidate's point-total in the "ABC methods" depends on how
overall-liked he is, not on how favorite he looks at one moment in
the count.

Borda has serious strategy problems, and most rank methods have
them to a somewhat lesser degree. There are some very good pairwise-
count methods, but only a few are as good as Approval in terms of
strategy problems.

The simplest is about the best, and is certainly the best public
proposal, since it amounts to only such a small change from Plurality--
the addition of one voting freedom, the elimination of the restriction
that we may only give a point to one candidate.

Mike Ossipoff

_________________________________________________________________
Join the world’s largest e-mail service with MSN Hotmail. 
http://www.hotmail.com