[EM] Misinformation in Approval, 3 candidates

[MIKE OSSIPOFF]

At least in 3-candidate examples, when Approval fails to elect the CW, 
Approval fails in favor of the middle. That isn't true of IRV, which tends 
to fail by jumping to an extreme.

Typically the voters will have the same information available to them. That' 
a reasonable assumption.

Typically it will appear, before the election, with 3 candidates, that one 
of the 2 extreme candidates will outpoll the other. The extreme preferrers 
who expect their extreme to be outpolled by the other will vote for Middle, 
in addition to their favorite.

I've posted this before, but say there are 3 candidates: Best, Middle, & 
Worst. I'd previously said "Favorite" intead of "Best", but "Favorite" could 
be construed as implying a plurality.

Say it appears that Best will outpoll Worst. Then the Best voters apparently 
have no need to vote for Middle. But say everyone's wrong about that, and 
that actually Wort outpoll Best. Believing that Best will outpoll Worst, the 
Best voter don't vote for Middle. But, believing the same thing, the Worst 
voters will vote for Middle.

Whichever extreme appears to be about to be ouitpolled by the other extreme 
will vote for Middle. Then, unless the oppoite extreme has a majoity, Middle 
wins.

But say that it appears that Worst will outpoll Best, and so the Best voters 
vote for Middle. And it turns out that Best would have had a majority, and 
would have won if his voters hadn't voted for Middle.  Then, in that 1st 
Approval election, Approval will have failed to elect the CW, due to 
misinformation. But the error was toward the middle.

So, when Approval errs, it tends to err toward the middle, whereas IRV tends 
to err by jumping to extremes. Obviously an error to the middle isn't as 
dangerous as an error to an extreme.

Pre-election polls to inform Appoval voting:

First, as I was saying, false answers aren't the problem with polls. The 
reliability of those who count and report the polls is the problem. But 
suppose that, in an Approval system, polls were being used officially or 
semi-officially, in such a way as to have verifiable ballots. The advantage 
of polls that sample a fraction of the electorate, as opposed to asking 
everyone,  would of course be a saving of money.

Our political polls, of course, ask the wrong question: Instead of asking 
people who their favorite candidate is, they ask whom the person would vote 
for if the election were held on that day.

In a system using repeated Approval polls to inform voting in the actual 
election, the 1st poll should at least encourge voters to just vote for 
their favorite. That isn't dishonest; it's the most useful information for 
the 1st poll to gather. Either it should be a 1st choice poll, or, if it's 
an Approval poll, people should be encouraged to vote only for their 
favorite. The only reason to use Approval in the 1st poll would be for those 
voters who reallly consider two candidates equally best, or so much better 
than the rest that their difference is negligible in comparison.

Subsequent pre-election polls would then use Approval, encouraging people to 
vote strategically based on previous results.

Of course, instead of going to the trouble of taking all those polls, it 
would be easier and better to just do the election (or at least the polling) 
by Condorcet. If polling were by Condorcet there'd be no need for more than 
1 poll. If the election were by Condorcet there'd be no need for polls.

Not that polls are actually needed anyway, of course. There's nothing wrong 
with starting out with a 0-info election, with subsequent poll-less 
elections informed by the results of previous elections.

About strategic insincerity in poll-answering: As I said, there's nothing 
dishonest about bullet-voting in the 1st poll. And it wouldn't serve much 
purpose in subsequent polls, where your favorite's winnability has already 
been revealed by the 1st poll.

With the 3-candidate election, in its pre-election polling, someone could 
say there'd be incentive to vote for the opposite extreme candidate, to 
deceive his voters into not voting for Middle when they need to. But if 
voting for him would make his voters believe that he outpolls Best, it could 
make other Best voters think so too. The deception doesn't sound like a good 
idea, if it wcld cause the other Best voters to needlessly give away the 
election.

Mike Ossipoff

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/