[EM] 01/05/03 - Olli Salmi and d'Hondt:

Donald Davison donald at mich.com
Sat Jan 4 22:10:58 PST 2003

01/05/03 - Olli Salmi and d'Hondt:

  ----------- Olli's Letter ------------
Date: Sun, 29 Dec 2002 16:47:55 +0200 (EET)
To: election-methods-list at eskimo.com
From: Olli Salmi <olli.salmi at uusikaupunki.fi>
Subject: [EM] Davison-SNTV

At 20:17 -0000 28.12.2002, Donald E Davison wrote:
> * Davison-SNTV is your next method of improved proportionality.  This is
>my variant of SNTV.  In this variant, instead of merely eliminating all the
>lower candidates at once, this method will eliminate them one by one
>according to the following rule:  `The candidate to be eliminated shall be
>the lowest candidate of the party with the lowest average votes per

How does this method differ from d'Hondt's rule which, when applied to
SNTV, is: `The candidate to be elected shall be the highest candidate of
the party with the highest average votes per candidate.'

Olli Salmi

- - - - - - - - - - - - - - - -
Dear Olli Salmi.

In a SNTV election method, you would be correct if you were to say that
d'Hondt will elect the same candidates as my policy of eliminating the
lowest candidate one by one of the parties with the lowest average votes
per candidate, but this is only true for SNTV, it will not be true in the
methods that use ranked ballots.

I prefer to have a consistent policy with all three methods, SNTV, Bottoms
Up, and Preference Votin/STV, therefore I also eliminate the lowest in

It is important in the two ranked ballot methods that the lowest candidates
be eliminated one by one so that their ballots can be transferred up and
have a bearing on who the final elected members are to be, otherwise there
will be no point in ranking ballots.

As far as the Party List method is concerned, d'Hondt should not be used,
but instead the voters should be allowed to rank candidates and/or
political parties in any mix and that the lowest candidates are to be
eliminated one by one.  In other words, I am suggesting that party list
method be turned into a ranked ballot method so that the voters will be
able to `enjoy' the benefits of ranked ballots.

The use of d'Hondt in any election is a joke, it is merely: `Much ado about
nothing'.  It is gobbledegook.  It does nothing to improve the election.
It elects the exact same candidates as would be elected by the policy of
assigning seats to a party according to `votes per seat' and then giving
the remaining seats to the largest remainders.

Anyone with any mathematical sense should realize that the party with the
most votes is going to have the largest quotients, so what, d'Hondt is not
telling us anything that we don't already know, just give out the seats
according to `votes per seat' and be done with it.  These so called
mathematicians try to con us into thinking there is some magic in their
silly creations.  The best that can be said for d'Hondt is that it will do
no harm, but that is no reson to use it.  It adds unnecessary mathematics
to an election method.

Below is text about d'Hondt being used in Macedonia:

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

For the first time this year, the parliamentary elections will be carried
out according to the proportional model. The territory of Republic of
Macedonia will be divided in six election districts, with approximately
equal number of voters. Twenty members of the parliament will be elected
from each election district.

The parliamentary seats will be allocated by applying the d'Hondt formula.

The d'Hondt method takes the votes obtained by each list and divides them
by one, two, three, four, and so on up to the number of seats to be filled
(twenty in this case).

The quotients obtained are ranked from the largest to the smallest, and
seats are allocated to the lists with the highest average. If two parties
have identical quotient, then the drawing of lots will decide the winner of
the mandate.

For example: If three parties are competing for 10 parliamentary seats in
one election district, and the party A obtains 80,000 votes, B - 60,000, C
- 50,000, then the number of votes of each party is divided by one, two,
three up to 10.

list  votes   1       2       3       4       5       ...  10
 A    80.000  80.000  40.000  26.667  20.000  16.000  ...  8.000
 B    60.000  60.000  30.000  20.000  15.000  12.000  ...  6.000
 C    50.000  50.000  25.000  16.667  12.500  10.000  ...  5.000

When the quotients are obtained, they are ranked and each party wins as
many parliamentary seats as its quotients are ranked among the top ten.

  rank     list       quotient
  1.         A          80.000
  2.         B          60.000
  3.         C          50.000
  4.         A          40.000
  5.         B          30.000
  6.         A          26.667

  7.         C          25.000
  8.         A          20,000
  9.         B          20,000
 10.         C          16.667
 11.         A          16,000
 12.         B          15,000
 ...        ...            ...

This means that party A will win four parliamentary seats, B and C - three
deputies in the parliament.

All 120 Members of the Parliament will be elected in just one election round.

up        © 2002 MIA - Macedonian
Information Agency

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Donald here: In the above example, d'Hondt gives the exact same results as
`votes per seat', but d'Hondt involves far more mathematics.

Votes per Seat method:
 * Total votes are equal to 190,000:
 * Votes per seat will be equal to 19,000 for this ten seat example:
 * A Party with 80,000 votes is assigned 4.2105 seats:
 * B Party with 60,000 votes is assigned 3.1579 seats:
 * C Party with 50,000 votes is assigned 2.6316 seats:
                              Total     10.0000 seats

C Party has the largest remainder, therefore it receives the tenth seat.

"This means that party A will win four parliamentary seats,
 B and C - [will win] three deputies [each] in the parliament."

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

More information about the Election-Methods mailing list