[EM] The Golden Rule of Preferences:

I Like Irving donald at mich.com
Thu Apr 5 07:31:15 PDT 2001

- - - - - - - - - - - - - - - - - - - - - - - 04/05/01
Greetings List,

I, Donald, wrote:
>      What is wrong here is that Bucklin violates the Golden Rule of
> Preferences, that is, later preferences are not to harm nor help earlier
> preferences.

Martin Harper <mcnh2 at cam.ac.uk> replied:
"And what basis is this golden rule coming from? What makes it golden? What
makes it better than an alternate rule: "earlier preferences are not to
help or harm later preferences". Or, say, criteria such as monotonicity?"

Donald: So, the Golden Rule upsets you, I should think it would, being as
you are a promoter of Approval Voting, a method in which all the lower
preferences will harm earlier preferences.

I recivied the Golden Rule from Stephen last month and have been repeating
it whenever it fits into a discussion.
It's a good rule, you would do well to learn it and abide by it.

Donald Davison,

 ------------ Forwarded Letter -----------
From:  Stephen
To: "'donald at mich.com'" <donald at mich.com>
Subject: RE: Irving and Robert's Rules of Order:
Date: Mon, 19 Mar 2001 16:38:16 +1200

Hi, Donald--

Re:  breaking ties (see under).

There can be situations where your current top choice and your next top
choice are not the tied candidates, and where your later preferences can
defeat your top choice.  If your top two choices are candidates C and E, and
candidates A and B are the tied candidates, your third preference for A
could help A survive, but later, when A is elected or excluded, his / her
surplus / votes could transfer to candidate D to help elect D over C.  Had
you not expressed that later preference for A, B might have survived
instead, and when B was later elected / excluded, his / her surplus / votes
could have transferred to C to help elect C over D.

The main argument against resorting to counting lower choices to break ties
is that once voters find out that there are circumstances whereby later
preferences could help defeat earlier preferences, they will be inclined not
to express later preferences.  Your preferred method of breaking ties breaks
the "golden rule" of STV and PV / AV / Irving (even if only rarely), that
being, of course, that later preferences cannot help / harm earlier

The whole purpose of promoting STV / Irving is to enable voters to obtain
equal and effective representation.  The "golden rule" gives  voters
complete confidence to express their true preferences (as far as they feel
able to) and, by doing so, to have as much influence over the composition of
the elected body as possible.  If the "golden rule" does not hold fast under
all circumstances, voters no longer have that complete confidence.

Here in New Zealand, the counting rules pertaining to Meek-STV / PV
elections, as set out in the Local Electoral Bill (due to be reported back
from Parliament's Justice and Electoral Committee on 30 April -- see
www.parliament.govt.nz), provide for the breaking of ties by first resorting
to the "Ahead at first difference" rule.  If that does not break the tie,
then a pseudo-random process (based on the Wichmann-Hill pseudo-random
generator [algorithm AS 183 -- see the book, Applied Statistics Algorithms,
Eds. Hill and Griffiths, 1985, Ellis Horwood Ltd, pages 238-242]) is used to
choose one of the tying candidates.  (The votes in STV / PV elections in NZ
will be counted by computer.)

Doing things this way enables ties to be broken with some basis in the
ballot papers (that has no impact on the sanctity of the "golden rule"), but
if that fails, to go to a pseudo-random process that will always break them.
This seems best and fairest to me.

Best regards

PS.  If you want to post this to the Instantrunoff site, by all means do so.

-----Original Message-----
From: donald at mich.com [mailto:donald at mich.com]
Sent: Monday, 19 March 2001 00:22
To: Instantrunoff
Subject: Irving and Robert's Rules of Order:

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 03/18/01

      I was looking over some pages out of Robert's Rules of Order that
came to me via email, when I seen something I didn't like.


     There is a criticism of using the next lower choices to solve a tie.
It is possible for your next choice to help defeat your current top choice.
You can consider this to be a slight possibility, because in order for
this to happen to you, four conditions must occur at the same time:

       One: There must be a tie, which is a rarity and the rarity
            increases as the elections become larger.
       Two: Your current top choice must be one of the tied candidates.
     Three: Your next choice must be the other tied candidate.
      Four: Your top choice must lose the tie contest.

    It will be rare for all of these conditions to happen to you at one
time, but they will happen to someone in the event of a tie.
    If these conditions do pile up on you and you become one of these
someones, your next choice will go on to compete with the rest of the
candidates, but compare this to what happens when we use Robert's Rules,
both your current top choice and your next choice will be eliminated,
neither will go on to compete with anyone.
     This Next Lower Choices solution has this slight downside of your next
lower choice helping to defeat your current top choice.  I accept this
downside because it will be a very small downside compared to having both
of my choices eliminated.  After all, which is worst, having one choice
causing another choice to be elimiated or having both choices being
     Having your candidate involved in a tie will be rare. Add to that the
odds against your next choice being the other candidate in the tie, it
becomes something that may never happen to you, and if it does happen to
you, it'll be best to keep one of your top two choices still in the game.

    This solution can also be used if the tied candidates are in the
winning position, which will be a more democratic solution than the
Plurality solution of Robert's Rules or the random selection of tossing a
     Next Lower Choices solution is the best solution to use. 03/18/01

Donald Davison

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

More information about the Election-Methods mailing list