[EM] 12/17/02 - Tom Ruen's Thoughts on Approval and IRV in two letters:

[Donald E Davison]

12/17/02 - Tom Ruen's Thoughts on Approval and IRV in two letters:



  ------------- Forwarded Letter -------------
To: <instantrunoff-freewheeling at yahoogroups.com>
From: "Tom Ruen" <tomruen at itascacg.com>
Date: Tue, 10 Dec 2002 03:43:03 -0600
Subject: [IRV-freewheeling] "IRV in action" and "The value of a vote"

Dear IRV supporters,

Part I: Example small IRV election
Part II: Supporting argument for IRV over Approval/Ratings on the grounds of
being less demanding to voters time.

Comments are welcome.

Sincerely,
Tom Ruen

**********

Part I: IRV in action

I ran an IRV election in my Toastmaster club tonight, voting for best
Tabletopics speaker. Usually my club doesn't vote on best speakers, but many
clubs have weekly voting to encourage speakers to try their best. Usually
votes are done by plurality, at least I've never witnessed any other method
in the two clubs I've been a member. Being there can be 6 speakers or more
plurality often can fail a majority opinion.

So this meeting I called for the election and here's the results. In this
case there was one strong candidate, although it is still interesting
example.

We had eight voters and 6 speakers (candidates).

IRV count:
Round 1: George: 4, Tom: 2, Suzanne 1, Brian: 1
(Implicit elimination of Brenda and Sharon with no votes.)
Eliminate Suzanne and Brian (A slightly dubious step, except that neither
were ranked next on the other single vote ballots and thus the either would
follow the other in elimination.)
Round 2: George: 6, Tom 2
Winner George with 75%

Oops, for the pleasure of our idealist, Donald Davison, let's finish the
process...
Eliminate Tom.
Round 3: George: 7 (I suspect the only exhausted ballot, with 3 solid
choices offered, was from our ever humble George)
Winner George with 100%

And for curiosity it is always fun to look at an approval vote, adding all
the ranked candidates. In this case it shows only George was ranked on
enough ballots to have a chance to win a majority!

Approval vote: (Counting all ranked votes equally)
George: 7
Tom: 3
Benda: 3
Brian: 2
Sharon: 2
Suzanne: 2

Note - I don't offer an Approval count to suggest it has any meaning for
determining a winner, only that it does suggest the relative strengths of
many candidates in a quick and easy form to see.

******************

Part II: The Value of a vote

The above election made me think about recent discussion over the value of a
vote and how much a voting method demands from voters.

I notice that for myself, spending 60 seconds to choose between 5 good
speeches (humble speakers don't vote for themselves), I believe my vote was
effectively randomized by a effect I'd call "whim of the moment." That means
if you asked me ten times in a row, erasing my member of my thought process
in previous voting, I would offer significant variation in my preferences
each time.

There may be a number of conclusions possible from this acknowledgement. One
might be that elections that don't matter much will always have this, and it
means nothing at all. Another more dangerous conclusion might be that any
method is as good as any other.

For example, my approval count might suggest someone say "Hey - approval
does pretty darn well without much fuss at all!" And of course we might take
that seriously and decide in an election that there is no issue of power,
only member recognition, that we should expand approval into a ratings vote
which will probably do even better to quickly determine the most popular
candidate.

And in elections like this that don't much matter, I would have no objection
to Approval or Ratings to determine the winner. However when I return to my
issue of accuracy and stability of vote, I must ask what is the simplest
method that takes the least effort for voters - the less you ask of voters,
the more likely they will be up to the task asked of them, especially on a
trivial result like a weekly best speaker from a small club. We might ask
voters to rate all candidates on a 1-10 scale, but that is asking more from
them than ranking a few top choices.

I conclude that IRV is a superior method to Approval, even in a simple fun
election.

Virtues of IRV:
1. It is easiest to vote. (Given a desire for a majority winner)
2. It is most accurate of voter intent, only using more random lower choices
as a last resort.
3. It is fast and easy to count.

Any voting method that has a questionable commitment from voters to
thoughtfully consider all the choices and make a sound decision, should
focus on minimizing the demands on voters, minimizing the failures of their
incomplete analysis of their options. Single vote methods best satisfy this
in my mind because it minimizes (and equalizes) power from each voter.

Ranking certainly has some "randomness" to it since the voter must make a
decision over nearly equally good options, but in the end, with the only
other option as abstaining, making a top choice can be done by voters and
voters will not be hurt if they are unable to well consider lower
preferences.

I say for me there is no fundamental flaw of power of Approval or Ratings in
terms of unequal power. The failure of Approval for me is that it demands
more thought from voters. I support IRV over Approval because it best
protects voters from themselves. Voters won't "overvote" in IRV simply
because they vote too quickly.



  ------------- Tom Ruen's Second Letter -----------
From: "Tom Ruen" <tomruen at itascacg.com>
To: "Donald E Davison" <donald at mich.com>
Subject: Re: 12/12/02 - Clarification needed on `overvote':
Date: Thu, 12 Dec 2002 17:18:02 -0600

Hi Donald,

I took the term "overvote" from a plurality election to mean there was more
votes cast than seats, and the vote is considered spoiled and not counted. I
remember news articles about "overvotes" in Florida and that overvotes that
included Gore far outnumbered overvotes that included Bush.

I'd call bad approval "overvoting" to mean offering extra "pity" votes for
preferences lower than would have been voted for in a single-vote election.
(Good approval "overvoting" is when one would have abandoned a weak favorite
anyway in a single-vote election.)

Overvoting is most dangerous when there are MANY choices (and minimum
polling knowledge) and voters don't have time (or aren't willing to make
time) to make a proper judgment over a threshold for support. A single vote
would have been easier to judge because there is less freedom to vote badly.
Ideally if you want to maximize "informed" votes, single-vote methods should
do better.

Tom






----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em