[EM] Pseudo-election reform in California

[Gervase Lam]

> From: bql at bolson org
> Date: Tue, 1 Jun 2004 13:43:27 -0700 (PDT)
> Subject: Re: [EM] Pseudo-election reform in California

> On Tue, 1 Jun 2004, Dr. Ernie Prabhakar wrote:
> > I agree that Plurality in the first round suffers from all the same
> > problems as Plurality in the final election.   So, my question is --
> > if people *want* a two-round system, what is the most efficient
> > election method to use?  I think Ranked Ballots are probably optimal,
> > but Condorcet is designed for single-winner elections. Is Condorcet
> > still the best way to determine the optimal dual-winners, if the goal
> > is to reflect the diversity of the electorate (vs. just the two most
> > acceptable to everyone)?  Or would some form of IRV or STV accomplish
> > that goal?
> >
> > Put another way -- I'm trying to figure out the optimal way to help
> > these reformers achieve *their* goals, rather than trying to get them
> > to -change- their goals.   Any suggestions?
>
> The Primary is equivalent to a multi-seat election (2 seats). It should
> be a proportional representation system. A second round with a clone
> candidate in it would be terrible. The largest minority would be angrily
> disenfranchised, I expect.

I initially disagreed with this.  Why not go for a 2 Round Cumulative 
Approval vote?

The problem with 2 Round Cumulative Approval Voting, as discussed on the 
list in the past, was that it is possible for the largest voting faction 
to approve their favourite and then approve turkey candidates.  The other 
voting factions would then basically have a choice between a candidate 
that they most probably hate and a turkey.

A similar thing can be done with plain rank voting methods.  For example, 
for Condorcet, just rank your favourite first and turkeys second, third, 
fourth, etc...  Therefore, having this in the first round is not a good 
idea.

One obvious way around this problem is to decay the weighting of the 
ballots that have voted for the winning candidate.  This is what 
Proportional Approval Voting (PAV) does!

I think Sequential PAV is more appropriate in this situation than the 
Non-Sequential version of PAV is appropriate.  One reason is that it is 
easier to count.

The other reason is that (Non-Sequential) PAV has (only slightly?) more of 
a tendency to yo-yo between left and right wing candidates if the 
electorate and therefore the ballots are strongly polarised to the left 
and right.  Although this is appropriate for a multi-seat election, we are 
not dealing with a multi-seat election here.

In other words, (Non-Sequential) PAV may not choose the Approval winner.  
In comparison, Sequential PAV always chooses the Approval winner, which in 
this case, I think is appropriate.

See:
<http://lists.electorama.com/pipermail/election-methods-electorama.com/2002-February/007366.html>
and
<http://lists.electorama.com/pipermail/election-methods-electorama.com/2002-February/007381.html>
for details, which includes details on why it might be best to use d'Hondt 
weightings, even in this case.  However, I don't know whether the 
weightings are strong enough to stop the ballots having too much influence 
in choosing the next winner.

To a certain extent, I think Sequential PAV in the first round and 
Condorcet in the second would not be bad.  The first round could be seen 
as a more sophisticated version of candidates getting enough signatures to 
qualify for a ballot.  It also acts to reduce one of the cited problems 
that Condorcet has of electing a candidate who has 'weak' support.

However, I think Sequential PAV in the first round followed by a top two 
run-off is probably better purely because of the fact that it might be 
easier to "sell".

Thanks,
Gervase.