One last spin around the Livejournal elections. In comments to our previous posts, Mr. GB asked,
How many votes are by second and third accounts, especially RPG accounts? How many of them have separate unique email addresses? There could be a lot of discounted invalid ballots, and I suspect that'll hurt Lego more than anyone else.
This is a valid point, and we kept records of the last published raw figures to determine precisely if this happened. Our gut feeling was that, even if validation took place after the count, it would not alter the balance of percentages significantly, or (somewhat less likely) help the third candidates. In the event, it appears that vote validation was conducted during polling, which is the sensible way to do things.
I just went and redid my vote to remove anyone not a front runner.
G-g-g-g-g-g-g-g-g-go!
Ahem.
The list of parties and their apparent candidates was:
- English Democrats: Cambler, Jameth, Jette, Lizyd, Mrflagg. (7530 first preferences)
- Fandom: Daniidebrabant, Imc, Legomymalfoy, Qfemale, Randomposting, Sollitaire, Tango, Vichan (11,920)
- LJ United: Lordandrei, Squeaky19 (1446)
- Fandom and LJ United: RM (2826)
- Independents: Deathboy, Fuzzface00, JJmaccrimmon, Jmaynard*, Lotcosmonaut, Primitivepeople, Twirlandswirl* (2153)
* -The chance to discern party information about these candidates was lost by their simultaneous elimination.
To determine these loose - and sometimes unintended - party affiliations, we've started with the groups mentioned on Tuesday, and worked backwards, determining the largest group receiving votes as candidates were eliminated. Only when this largest group is clearly (in effect, 15%) larger than any other group, do we put them in a party.
Our model correctly predicted the last seven candidates, and (other than incorrectly guessing the near-tie between Squeaky and Vichan for 5th place) their elimination order. It was slightly out in the raw votes: both Legomy and RM had outperformed our model by 600 votes; the rest of the last six were between 100 and 200 votes adrift.
By the last three, RM was polling over 1000 ahead of our prediction, suggesting that she was the compromise candidate. In the end, RM's vote broke 3% to Jameth, 46% to Legomy, 51% NENT. Though Legomy finished 309 votes short of a formal Quota, it's almost certain that Jameth's pile contained enough.
A majority of votes from the elimination of Cambler (7th), Deathboy (6th), and RM (3rd) were non-transferrable. It would have been acceptable to eliminate Squeaky, Vichan, and RM in one fell swoop, as the three could not have combined to beat Jameth in second.
There was also an election amongst the Cyrillic userbase (which still includes all customers from the Baltic states. Great way to offend; it's as if 1991 never happened.) Party lines here were much weaker, and we could only put half the candidates in one camp or another. Almost 30% of the votes became non-transferable, further showing the weakness of the truncated AV model used here. In the end, Zafar F Khashimov, a 40-year old from Moscow, emerged at the top of the ballot; we suspect that he would not have secured the backing of a majority of the voters. With their usual attention to detail, Livejournal has released no biographical details of Mr. Khasimov. Anyone would think they didn't want to be associated with a pot-smoking liberal hippy.
In the interests of transparency, Livejournal should have disclosed that it modified the code used to process the count about four hours before the polls closed. In our view, the changes were minor, and did nothing to alter the integrity of the poll. Other readers' opinion may vary.
And finally, the turnout. Across the two votes, 3%. It's as if the whole of Wales were to be represented only by the opinions of the people of Anglesea. Still, at least the position holds precisely no power or importance.
