[EM] Anyone got a good analysis on limitations of approval andrange voting? (long)

[Raph Frank]

On Tue, Nov 10, 2009 at 8:18 AM, robert bristow-johnson
<rbj at audioimagination.com> wrote:
> i will say this: even though it is prohibited in the present IRV that
> Burlington VT has, there is no reason that ties should not be allowed in any
> ranked ballot.

It is unclear how this should work with IRV.

My personal preference is that both votes would count at full
strength.  The other possibility is that they are divided equally
between all remaining candidates.

In both cases, the count is more difficult.

> is that putting it in an accessible context?  so then with Approval voting,
> for sure this grandma marks "X" by her grandson's name.  but does she or
> doesn't she Approve her good ol' incumbent?  Approval doesn't let you mark
> it "Approve except in the race with my top choice."

However, if you know who the top-2 are, then this isn't a major problem.

The point with approval is that it tends to converge to a condorcet
winner, or a candidate who is almost as good.

It might even be better at finding honest condorcet winners than an
actual condorcet method, assuming that the voters vote tactically in
both cases.

> i mean isn't that the essential flaw?  e.g. the flaw with the Electoral
> College is that sometimes it elects the wrong candidate.  it does pretty
> good when it selects the same winner as the popular vote, but when it
> disagrees with the popular vote it *never* creates more legitimacy or
> confidence in the election results.  so then why have it?  what good is it?

It would have probably been better if they set it up so the college
actually meets, but then add a requirement that the majority of the
electors must agree.

If combined with PR at the state level when electing electors, then a
candidate who has the support of the majority of the population should
end up winnng.