Re: GNOME Board of Directors Foundation Elections Spring 2009 - Preliminary results

[Dave Neary <dneary gnome org>]

Hi Tobias,

Tobias Mueller wrote:
> These results can be challenged by sending an e-mail to
> elections gnome org  The challenges have to be sent before Tuesday,
> June 30, 2009, 23:59 UTC. Please note that these results should not
> be considered final until any such challenges have been resolved.

Challenge!

You just announced the results based on first-past-the-post, when the 
elections were to be run using preferential voting, with single 
transferable vote and fractional surplus transfer.

Srinivasa and Diego get elected in that system instead of Jorge.

> Candidates in order of votes received, with affiliations:
>
>       Vincent Untz           (60 votes) - Novell, Inc.
>       Behdad Esfahbod        (56 votes) - Red Hat
>       GermÃn PÃo-CaamaÃo     (20 votes) - None
>       Brian Cameron          (19 votes) - Sun Microsystems, Inc.
>       Jorge Castro           (14 votes) - Canonical Ltd.
>       Lucas Rocha            (13 votes) - litl
>       Srinivasa Ragavan      (10 votes) - Novell, Inc.
>       Diego Escalante Urrelo ( 8 votes) - None
>       Hubert Figuiere        ( 7 votes) - None
>       Og Maciel              ( 4 votes) - rPath Inc
>
>
>
> If the results are not challenged, then the elected directors will be:
>
>       Behdad Esfahbod
>       Brian Cameron
>       Diego Escalante Urrelo
>       GermÃn PÃo-CaamaÃo
>       Jorge Castro
>       Lucas Rocha
>       Vincent Untz
>
>
>
> Some figures about the votes: there were 382 registered voters. 213
> voters sent valid ballots. Two votes were not valid or empty.
>
> The results can be found at
> http://foundation.gnome.org/vote/results.php?election_id=13
>
> A list of all votes can be found at
> http://foundation.gnome.org/vote/votes.php?election_id=13


The table on this results page shows the results with random transfer 
STV. When this was discussed over at Maemo, it was proposed & agreed 
that we should use the more accurate fractional transfer STV, since 
running random transfer STV several times can give different results.

If we use the mathematically more correct fractional transfer, the 
results will stay the same no matter how many times you count.

The first method treats votes as pieces of paper, and in the cases where 
candidates have a big surplus, this can make a difference.

We can tell from the count that of the 60 people who voted for Vincent 
in the first count, the number 2 preferences (or number 3, if Behdad was 
number 2) looked like this:

Brian: 13
Diego: 3
German: 7
Hubert: 2
Jorge: 0
Lucas: 26
Og: 3
Srinivasa: 6

33 of the 60 votes get transferred.

In fractional surplus distribution, each candidate gets N*33/60 
transfers - if we ever have to distribute a surplus later, these votes 
are weighted to be worth roughly half a vote.

In random surplus distribution, we canculate how many votes each person 
gets transferred, rounded:

Brian: 7
Diego: 1
German: 4
Hubert: 1
Jorge: 0
Lucas: 16
Og: 2
Srinivasa: 2 (and seems like a bug, since Srinivasa should have gotten 3 
or 4)

The sum total is the same - we transfer 33 pieces of paper. But each of 
these have the weighht of a full vote, and the actual votes we transfer 
are randomly chosen among all the votes that had 1. Vincent, 2. Lucas 
(for example). And since they're chosen randomly, it's possible that the 
#3 or #4 preferences on the transferred ballots aren't representative of 
the population.

In any case, I'd like to have fractional transfer STV used to count, 
rather than random transfer.

Cheers,
Dave.
--
Dave Neary
GNOME Foundation member
dneary gnome org