[EM] 01/31/02 - Two new design features for STV:

[Donald Davison]

01/31/02 - Two new design features for STV:

Greetings List members, recently I wrote in reply to Adam's letter:

>Donald:  Yes, it has been proposed before to allow people to vote for
>candidates and/or parties in any mix.

Adam replied: "...I didn't see any specific mention of this in the
following list..."

Dear Adam,

Donald here: I did have a proposed Plan at the top of my list of multi-seat
methods, but took it off about a year ago, pending a debate I was having
with myself.  I was not satisfied with how I was tallying the mix of
candidate and party votes.

It is one thing for you and I to say that STV should have this positive
design feature of allowing people to vote for candidates and/or parties in
any mix, it is another thing to say how we are going to tally the ballots.
For example:

A)  Do we give the party votes a list of choices like a Closed Party List?
    No, because you and I do not like Closed Party List.
B)  Do we give the party votes a list of choices like an Open Party List?
    This is better, but the lowest candidate will receive little if any
benefit from the party votes.  It can be argued that the party votes were
intended to help every candidate of the party as much as possible without
harming any candidate.
C)  Do we divide the party votes in proportion to the votes of the candidates?
    This is also better but still not good enough, the lower candidates are
still not being protected enough.  Protection means that every candidate of
the same party, is to be helped to avoid as many elimination cycles as
possible.
D)  Do we divide the party votes equally between all the party's candidates?
    This is the option I was using in my proposed Plan, but I took the Plan
`off the table' because the math got too messy.
E)  I now have a new option, which came about because Tom Ruen wanted to
use vote splitting in a STV election so that he could protect two
candidates.  I told him vote spliting would do a poor job of protecting his
two candidates, so I invented for him a new elimination rule for STV that
would do a very good job of protecting, as much as possible, all the
candidates of a faction/party.  That rule is as follows:

      "The candidate to be eliminated shall be the lowest candidate
       of the faction with the lowest average votes per candidate."

It did not take long for me to realize that this rule would make the math
much easier for my election Plan that mixed candidate and party choices.
With this rule the party votes would not need to be divided between the
party candidates.  When it came time to eliminate a candidate in the entire
election, the party votes would be included in calculating the party's
average votes per candidate (at the `side bar' of course).  In this way the
party votes will be helping all the party's candidates to avoid as many
eliminations as possible, but there will be no transferring of party votes
back and forth - less math.

Also, we can expect that more than fifty percent of the voters will be
voting for party, which inturn means less surplus votes, that also means
less math.

So, currently, that is where I am at in this project.  I see partisan STV
with two new design features:  One, allowing choices for candidates and/or
parties in any mix.  Two, using my new elimination rule for candidates.  As
of now, I see no harm that can be caused by these two new design features,
only good.

What does the public want?

I say the public wants to be able to vote for candidates and/or parties in
any mix.
I say the public also wants all their candidates to be protected as long as
possible from elimination even if it means using both party votes and the
votes of the higher candidates of the party (only temporally of course).

The elimination rule has another advantage.  In the event the public wants
the option of splitting votes, my new rule will prevent that option from
harming any candidate.

Regards, Donald

Below I forward Tom Ruen's post of January 22 which this list has not yet seen.
I also forward my reply to Tom's post - enjoy.

- -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
To: <stv-voting at yahoogroups.com>
From: "Tom Ruen" <tomruen at itascacg.com>
Date: Tue, 22 Jan 2002 22:36:35 -0600
Subject: Re: [stv-voting] 01/20/02 - FWD - STV for corporate board elections:

Dear Joe and Others,

I agree that following standard rules is a good idea. Still, I must consider
some questions.

My current most important question arises from the fact that Cumulative
Voting (CV) allows voters to explicitly divide their vote while STV only
allows rankings. Some voters might not want to limit their first vote to one
candidate. This might be strategically valuable if a voter is supporting 2
candidates and want to minimize the risk that one or the other will be
eliminated.

On a stock-based vote, I would consider allowing voters the option to use
two or more ballots where the sum of votes among the ballots add to their
full votes. Example, if I had 300 votes, I might want to put 150 on one
candidate and 150 on another.
Instead of deciding between:
300: AB...
or
300 BA...
I could have two ballots:
150: AB...
and
150: BA...

I see no reason to disallow this option for voters. I believe it could
potentially affect the elimination order of candidates since that can be
very close. The longer I can keep a set of favorites in the running, the
longer I can hope transfers can still elect both of them.

This might not be thought of for political elections with thousands of
voters, but is quite natural for voters moving from CV elections. This
option allows stubborn voters to continue with a CV type vote if they like.
And such voters still gain by adding lower rankings on each.

Does anyone have any reason to disallow one voter from splitting his or her
vote on more than one ballot with different rankings?

I've heard the argument (from 2 people so far) that offering this power to
voters shouldn't affect the election, but that's not what I'm asking. I want
to decide whether to allow voters this option if they choose it.

Thanks for listening,

Sincerely,
Tom Ruen

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
To: stv-voting at yahoogroups.com
From: donald at mich.com
Date: Sat, 26 Jan 2002 06:00:21 -0500
Subject: [stv-voting] 01/26/02 - Let's Fiddle with the rules of STV:

Dear Tom,

The greatest value of vote splitting may be in public relations.  If most
of the public likes vote splitting, for whatever reason, then the public
should be more inclined to adopt STV.  If so, I will support optional vote
splitting because I am in favor of more STV elections.

Having said that, I also believe vote splitting is kind of benign, that is,
it will help your lower candidate sometimes and it will harm your higher
candidate sometimes, but most of the time it will merely be a placebo.  The
option of splitting votes will do little to implement the policy of
protecting candidates of the same faction.

Only in the rare case in which candidates A and B are neck and neck after
the transfer of surplus votes will the equal division of your votes between
the two protect each equally, but with 400,000 votes and only six candidates,
we can expect thousands of votes of seperation between the candidates.  A
mere 300 votes will not be of much help.

A slight improvement would be for you to depend on a pre-election poll.
Once you know which of your two candidates is the lower one, you can then
place all 300 votes on that candidate, but even this will not solve your
dilemma.  This candidate may still be a thousand votes lower than your
other candidate.  You must work your policy in a larger context.

The success of candidates A and B depends mostly on the entire voting bloc
that is supporting these two candidates.  Suppose sixty percent of this
voting bloc would vote for candidate A as their first preference and B as
second, leaving forty percent voting BA.  But, let's say, they listen to
you and they agree to divide the votes equally between A and B.  Suppose
the policy is working so far in that your two candidates are able to avoid
the first few eliminations.  And, suppose candidate A has received 1000
votes in transfers and candidate B has received 1005 votes.  If this
addition of 2005 votes is enough to give the bloc two quotas, then your
plan has worked - both A and B are elected.  But, suppose the bloc has less
than two quotas and it is now facing the time for one of its candidates to
be eliminated.

Which one is to be eliminated?  According to the current numbers, candidate
A has five votes less than candidate B.  Are you going to eliminate
candidate A when he would be the leading candidate of the two if not for
your advice?  Your option of splitting the votes is causing a lower
preference to harm a higher preference.  In other words, if we are to
divide the votes equally between two candidates, but only one can be
elected, which is elected?


Allow me to fiddle with the rules of STV and come up with a much better way
to implement your policy of helping both your candidates.  Instead of
eliminating the lowest candidates during the runoff cycles of STV, we use
the following new rule:

      "The candidate to be eliminated shall be the lowest candidate
       of the faction with the lowest average votes per candidate."

This rule will give the best possible protection to every candidate of
every faction thru every runoff cycle of the STV election.  Mathematically,
this rule will enforce your policy, but the question remains: "Should your
policy be allowed in STV?"

Side Note:  An advantage of this rule is that it will make it easy for a
partisan STV election to allow the people to vote for candidates and/or
parties in any mix.  The averaging will always divide all votes equally
between all the remaining candidates of a party without the need to split
up any of the party votes.

Regards, Donald Davison