Re: Election results

Dear Antti-Juhani Kaijanaho,

you wrote (7 Aug 2007):

> I'd be interested in how Schulze's STV method (a generalisation of the
> Debian method for a multiple winner situation) handles this election.

The Schulze STV method would have chosen

B: Luk Claes
D: David Graham
E: Bdale Garbee
F: Joshua D. Drake
K: Martin 'Joey' Schulze
L: MJ Ray

I have attached the output file of my program.

Markus Schulze

Markus Schulze <markus(dot)schulze(at)alumni(dot)tu-berlin(dot)de> wrote:
> The Schulze STV method would have chosen [...]

To summarise the interesting (IMO) results posted so far:

Would have been elected under all systems:
D. David Graham CSPI CBorda CRP SSTV AV STV FPTP
F. Joshua D. Drake CSPI CBorda CRP SSTV AV STV FPTP
E. Bdale Garbee CSPI CBorda CRP SSTV AV STV FPTP
B. Luk Claes CSPI CBorda CRP SSTV AV STV FPTP

Elected under some systems:
C. Joerg Jaspert CSPI CBorda CRP STV
H. Ian Jackson STV
A. Martin Zobel-Helas CSPI CBorda CRP
L. MJ Ray SSTV AV FPTP
K. Martin 'Joey' Schulze SSTV AV FPTP

I replaced my earlier list position barchart in
http://mjr.towers.org.uk/blog/2007/spi#resultelections
with one showing all candidates' list positions.

Is it interesting that we elected the candidates who appear in list
positions 9-13 least? There seem to be some odd interactions between
votes for H and K in particular, and maybe C and L, which may explain
some of the different results. Is SPI actually election-by-veto?

Is it interesting that we elected ABCDEF? Was there an alphabet
effect? ;-)

Regards,
--
MJ Ray - see/vidu http://mjr.towers.org.uk/email.html
Experienced webmaster-developers for hire http://www.ttllp.co.uk/
Also: statistician, sysadmin, online shop builder, workers co-op.
Writing on koha, debian, sat TV, Kewstoke http://mjr.towers.org.uk/

On Thu, 9 Aug 2007, MJ Ray wrote:
> To summarise the interesting (IMO) results posted so far:
> [...]

It should be noted that the electoral system used is likely to change
voting behaviour, so this discussion can be fairly moot.

Let's say, to take an extreme example, that, because not everyone ranked
every candidate, we can count this election with Approval Voting, a very
simple system where you simply check off everyone you approve of. We count
it with that and here are the results:

93 * Candidate E "Bdale Garbee"
90 * Candidate A "Martin Zobel-Helas"
88 * Candidate L "MJ Ray"
88 * Candidate F "Joshua D. Drake"
88 * Candidate B "Luk Claes"
87 * Candidate K "Martin 'Joey' Schulze"
87 * Candidate C "Joerg Jaspert"
86 * Candidate D "David Graham"
84 * Candidate H "Ian Jackson"
84 * Candidate I "Christel Dahlskjaer"
83 * Candidate G "Rico Gloeckner"
83 * Candidate J "Richard Hartmann"
81 * Candidate M "Robert Brockway"

The results are, like all the other systems, fairly similar to the actual
results, with me off and MJ on, and Joey and Ganneff tying for 6th, but
voting behaviour changes with the system used. Comparing one election
system's votes to another election system's results is not really a valid
comparison.

> Is it interesting that we elected the candidates who appear in list
> positions 9-13 least? There seem to be some odd interactions between
> votes for H and K in particular, and maybe C and L, which may explain
> some of the different results. Is SPI actually election-by-veto?
>
> Is it interesting that we elected ABCDEF? Was there an alphabet effect?
> ;-)

This isn't terribly surprising. Putting your nomination in early helps
more people know about you, your campaign, your attitudes, and your
commitment, while putting in your nomination late, or as an overt reaction
to the existing candidates can be off-putting to voters. The former
suggests that you will not be active in board activity while the latter
suggests you may be too judgemental to make a good board member.

- -
David "cdlu" Graham - cdlu(at)railfan(dot)ca
Guelph, Ontario - http://www.cdlu.net/

David Graham <cdlu(at)railfan(dot)ca> wrote:
> It should be noted that the electoral system used is likely to change
> voting behaviour, so this discussion can be fairly moot.

To some extent, but if people know what they're voting under SPI's
system, and they're not tactical-voting (I agree with Wouter Verhelst
http://grep.be/blog//en/life/politics/voting_tactics_II?show_comments=yes
about tactical voting) - that is, that a vote of ABCDEF for candidates
A-F really does mean that they prefer A to everybody, B to C and so on
- then the votes would be the same for several systems.

Not Approval Voting, as you rightly point out, but STV, SSTV, AV and
FPTP are valid comparisons, aren't they? Maybe the other Condorcets
too? Do the SPI votes fail those conditions, or did I miss something?

[...alphabet effect]
> This isn't terribly surprising. [...]
> [... Late nomination] suggests that you will not be active in board
> activity while [overt reaction to the existing candidates] suggests
> you may be too judgemental to make a good board member.

The other comments about early nomination are good points, well made.

This might not be terribly surprising to experienced SPI candidates,
but it did surprise me, as I'd not really thought of it like that.
Most elections I've voted in group and publish all nominations
together in one round. OTTOMH, I can only think of SPI and debian
doing it like this, where you publicise your own candidature without
restriction on campaigning. I guess this will increase as more voting
moves online and more society members communicate with each other
directly, rather than through moderated mailings, so we're blazing a
trail again!

There are many reasons possible for a nomination date, so relatively
late nomination doesn't suggest future inactivity to me (as long as
the nomination is well before the deadline!), but I think very early
nomination means you were going to stand no matter what. That may or
may not be a good thing.

I'm used to candidates reacting to each other overtly from debian's
debate/rebuttals system, so I didn't expect that to be a problem in
SPI elections. I may have misjudged that one.

SPI election results were ABCDEF, ABD, AB, ABD, GBE. (That G who
totally flouts the alphabet effect is Bruce Perens. The unelected
A was Scott Dier and C was one David Graham.)

Finally, a list of the last three year's DPL nomination dates is
http://lists.debian.org/debian-vote/2007/07/msg00178.html
in summary:
2007 - nominations 4th-24th, winner nominated on 14th (11th day);
2006 - 5th-26th, winner on 23rd (19th day);
2005 - 7th-27th, winner on 27th (21st day);
and before that:
2004 - 7th-28th, winner on 15th (9th day);
2003 - 24th-14th, winner on 13th (21st day);
2002 - 6th-27th, winner on 13th (8th day);
2001 - 24th-13th, winner on 24th (1st day);
2000 - 9th-31st, winner on 26th (17th day);
1999 - predates the archive I'm using.

Would anyone like to suggest why SPI has a different nomination-date
effect to debian? (Be gentle...)

Thanks,
--
MJ Ray - see/vidu http://mjr.towers.org.uk/email.html
Experienced webmaster-developers for hire http://www.ttllp.co.uk/
Also: statistician, sysadmin, online shop builder, workers co-op.
Writing on koha, debian, sat TV, Kewstoke http://mjr.towers.org.uk/