[EM] simple methods' merit limit

[Forest Simmons]

Here's an idea for a fairly simple "non-simple" method:

Each voter is issued an Approval/Plurality style ballot.

The voters cut the ballots up into strips with one candidate per strip. 

Each voter submits one strip per round in the order he/she considers
advantageous. 

Between rounds the results are made public.

The winner is the candidate with the greatest approval.

[End of description of the method]

This method could be called Slow Approval, Careful Approval, Fully
Informed Approval, Gradual Approval, etc.

[Bart might not like me to use "Approval" as part of the name, but leaving
"Approval" out of the name would be claiming more originality than I
deserve credit for.]

The end result is that each voter has expressed approval or lack thereof
for each candidate, so the method is Approval with a novel way of
collecting the ballot information.

What advantage does the gradual ballot submission have?

Each round serves as a binding poll to inform the voter choices for the
next round.

That makes the method harder to manipulate by bogus polls, etc.

Consider the case of three candidates:

Suppose your sincere preference order to be   Nader >> Gore > Bush .

In the first round you would submit the Nader strip with approval.

In the second round you would submit the Bush strip w/o approval.

After the results of the second round are announced, all of the other
voters would know your preference order, except for the strength of
preference (indicated by the double inequality above).

Similarly, you would know their preference orders.

You would be informed well enough to make an intelligent choice about
whether or not to approve Gore on your final strip.

If any of the three candidates had a majority of first place preferences,
then that candidate's supporters would know it and would (bullet) vote
accordingly, assuring victory for their favorite.

So with this method of balloting, we find ourselves close to the perfect
information case where the Majority Criterion is (strategically)
satisfied.

Similarly, the Condorcet Winner is more likely to be chosen by this method
of balloting than by the all-at-once approval ballot method. This is true
because gradual balloting moves us towards the "perfect information case",
and the CW is more likely to win in the perfect information case than in
the zero information or bogus information cases.

Of course, this would work best in small applications like committees for
text book selection or for hiring new faculty.

But consider the November 2000 presidential election.  The interest
generated by that for two months would have sustained an election of this
style, and probably with better voter satisfaction.

Forest

On Wed, 28 Nov 2001, Forest Simmons wrote:

> 
> 
> On Wed, 28 Nov 2001, MIKE OSSIPOFF wrote:
> 
> <snip>
> 
> > 
> > So maybe the simpler non-simple methods should be considered as
> > public proposals.
> > 
> 
> They could certainly be used voting in committees.
> 
> For large scale public elections we would probably have to limit the
> elections to two rounds, as is done in most actual runoffs.
> 
> However, a more complicated method could be used by the proxies in an
> election where the voters just designate their favorites to be their
> proxies.
> 
> If this is done correctly, the voters will have little incentive to
> designate anyone besides their true favorites as proxies. This gives the
> proxies the advantage of knowing their opponents' first place strengths
> right off of the bat.
> 
> To Be Continued
> 
> Forest
> 
>