[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Vote analysis

Some quick analysis of the tally sheet:

First preference votes:

    152 Anthony
    122 Steve
     75 Andreas
     70 Jeroen
     48 Bill
      9 Ari
      4 Ted

     14 NOTA

(that counts 48 votes that ranked two or more options equal first
multiple times)

Using STV (aka IRV, which .au uses for its elections) instead of Condorcet
to get the result, gives me eliminations in the following order:

    8. Ted (4 votes)
    7. Ari (9 votes)
    6. NOTA (14 votes)
    5. Bill (57 votes)
    4. Jeroen (93 votes)
    3. Andreas (121 votes)

with the end result being me beating Steve by 6 votes, at 237 v 231.

That's not really legitimate STV since when a vote ranks two candidates
equally I count it towards both totals, which is why 231+237 = 468 which
is more than the total number of votes (421), but it's the best we can
do, I think.

So, by the looks of things, we get the same result with either
American-style voting (only the first ranked candidate counts) or
Australian-style (preferential, but eliminations rather than pairwise
comparisons). Approval voting (against NOTA) would result in:

    344 Steve
    339 Anthony
    321 Andreas
    319 Jeroen
    294 Bill
    158 Ari
     73 Ted

Similar result if you consider equal to NOTA as "approval", except that
in that case Jeroen and Andreas tie. Interestingly, approval voting
(against NOTA) would've given a different result last year too, with me
and Matthew Garrett beating Branden (though who exactly won depended on
which method of defining "approval" you choose). Though, again, I don't
really think people would've voted the same way if that was the system
we used. (cf http://lists.debian.org/debian-vote/2005/04/msg00039.html)

If I remember how my script works well enough to use it correctly,
the "clone" candidates (ones that a lot of voters rank next to each other)

    326 votes (77%), Ari and Ted
    269 votes (64%), Jeroen and Andreas
    260 votes (61%), Ari and Bill
    247 votes (59%), Jeroen and Steve
    244 votes (58%), Steve and Anthony
    228 votes (54%), Steve and Andreas
    220 votes (52%), Andreas and Bill
    217 votes (52%), Anthony and Andreas

This year, 338 votes were unique in meaning, with an additional 12 unique
in expression (eg, "1-------" vs "12222222"), leaving 71 voters who voted
exactly the same as someone else, and 83 who ordered the candidates in
one of the 36 orderings that more than one person used. The most common
orderings were:

      6 V: 22222221   (NOTA, everyone)
      4 V: 33313332   (Anthony, NOTA, everyone) 
      4 V: 22212222   (Anthony, everyone)
      3 V: 47213856   (Anthony, Steve, Andreas, Jeroen, Bill, NOTA, Ari, Ted)
      3 V: 33123333   (Steve, Anthony, Everyone)
      3 V: 22221223   (Andreas, NOTA, everyone)

No other ordering had more than two people vote for it. (You can do
the maths yourself, but that leaves 30 pairs of people who voted in the
same way)

How deeply did people express preferences?

      0 Didn't express any preferences (all options equal)
     16 Only had two different preferences (eg, 11111112)
     29 Three preferences
     50 Four preferences
     48 Five preferences
     50 Six preferences
     51 Seven preferences
    177 Full preferences (every option ranked differently)

Apparently two people voted the outcome we actually got:

    V: 37213845                 7ecd14e14664e97b71ed6fd6ae9e3289
    V: 36213745                 9df7140f924e1aba6d7022604a60db5c


Attachment: signature.asc
Description: Digital signature

Reply to: