Multi-winner Wikimedia Foundation election results and advice for the future

The results of the multi-winner Wikimedia foundation board election which I noted in STV (Single Transferable Vote) options for surplus transfer, elimination etc: Baas, Meek, Warren have now been posted and reported on:

along with an interesting commentary:

It notes just how close the race for the 2nd of two seats was. 40.52 vs 40.48 votes out of 122 ballots, or 0.03 %!
You can also see that some affiliates seem to have cast their ballot in unusual ways suggesting lack of familiarity with the rules in use (e.g. they seem more like scores than ranks).

As you can see from the ballots (all public and identified by affiliate) and the flexstv Python software used to tally it and previous elections, not only could a vote for any of three candidates have changed the results, but as it typical with ranked-choice elections, changes of other ranking also could have resulted in a different outcome. Thankfully, the second-place winner was also the Condorcet winner. I consider it unfortunate that the results were tallied as if all ballots actually ranked all candidates (via equal rankings for the unranked candidates). That makes the final results seem like everyone supported one of the two remaining candidates in the final round.

This election was on a very tight deadline and I got involved at the last minute, after decisions about voting methods and tools had been made.

I remain very interested in what voting method and tooling to suggest for the next such election (in 3 years), now that we have time to actually consider the matter with experts and the community.


What PR method or tweak would you personally suggest to them?

How is this possible? Are the voters given unequal weights from the outset?

Weights are all equal to start. One unusual and significant factor is probably the allowance for equal rankings and the way the code splits such votes equally among all remaining candidates.
Check out the spreadsheet on the Results page to see it evolve round-by-round, or try the code out.

@Sara_Wolf is holding a committee to try to evaluate various score PR methods. Perhaps the findings of the committee will be useful.


I think for this purpose, Proportional approval voting has a lot of simplicity and effectiveness to recommend it. I’m also very interested in (and participating in) the recommendations of the team that @Marylander noted. More info on that should appear at Proportional Representation - STAR Voting when it’s ready. But discussion here is welcome also!


That is susceptible to free riding, so what about my variation?

I think my variation always removes the same total vote weight (unless a candidate wins with less than a quota), where “total vote weight” means

total_weight = 0
for vote in votes:
    total_weight += vote.weight

Anything short of 100% (perceived or real) approval and this is also subject to free riding, albeit one with a certain strategic tension and impetus for honesty that can’t be fully analyzed in advance.