Wikipedia:Wikipedia Signpost/2009-12-21/Election report

On Thursday 18 December, the six scrutineers certified and announced the results of the two-week community-run election for the Arbitration Committee, the final dispute resolution body of the English Wikipedia. The nine new arbitrators, who will take up their seats on 1 January 2010, are Kirill Lokshin (72.9%), Fritzpoll (71.1%), Coren (70.1%), Mailer diablo (68.4%), Steve Smith (65.6%), SirFozzie (64.9%), Hersfold (64.6%), KnightLago (60.1%), and Shell Kinney (59.9%).

The Signpost congratulates the arbitrators-elect, and offers its thanks to the other 13 candidates for participating in the election. A formal announcement of the appointments by Wikipedia co-founder Jimbo Wales is expected soon. Just as for the previous round of appointments a year ago, a system of tranches will be used to minimise the variation of vacancies from year to year. The candidates with the five strongest votes—Kirill Lokshin, Fritzpoll, Coren, Mailer diablo and Steve Smith—will serve two-year terms expiring 31 December 2011; the remaining four successful candidates—SirFozzie, Hersfold, KnightLago and Shell Kinney—will serve one-year terms expiring 31 December 2010. The new ArbCom arrangement is illustrated in this graph.

Running of the election
The 2009 election was the first to use the SecurePoll system, and overall the election ran smoothly. There were only three minor glitches. First, due to confusion about the use of "00:00 UTC" in the software, the election was programmed to end midnight at the start rather than the end of 14 December. The problem was picked up during the weekend before the end of the election and fixed by MediaWiki developer Roan Kattouw. The second issue was that the scrutineering process took four days, more than the single day the election coordinators had advertised. During this time, the mostly good-humoured frustration of candidates and onlookers was clear on the election talk page. The third glitch was that due to an oversight, three duplicated ballots were not picked up, so 996 ballots were counted from the 993 unique voters. Happy-melon, one of the three election administrators, that the minimum number of votes that would be needed to alter the outcome was 13, "so there is no concern that this irregularity could prejudice the outcome of the election. To protect the privacy of the three votes, we have no intention of publishing a revised tally.... Had this irregularity been identified before the publication of the results, it could have been trivially rectified.... we also note the pressure that was placed on the scrutineers by members of the enwiki community to publish the results as rapidly as possible."

Comparisons with the 2008 election
This year, there were 993 unique voters, slightly more than last year's 984. About 40% of voters in 2009 also voted in 2008. In 2009, there were 22 candidates at the end of voting (one withdrew during the polling), compared with 28 candidates in 2008.

Three significant shifts in voting patterns occurred in 2009. First, the proportion of Neutral/abstain votes was dramatically reduced, from an average of 75.3% in 2008 down to 44.7% in 2009. In Figure 1, each of the 50 points represents one of the candidates (28 in 2008, above the line; and 22 in 2009, below the line). The blue points were successful candidates; the red points unsuccessful. Each point, then, shows two features of the votes for a candidate: (a) the ranking percentage, from left to right, which determined whether they were elected; and (b) the percentage of Neutral votes (equivalent to abstains) from bottom to top for each candidate as a proportion of the total number of voters (996 in 2009; 984 in 2008). Thus, candidates who fell towards the bottom-right received high ranking-percentages with low proportions of Neutral votes / abstains; and candidates towards the top-left gained low ranking percentages and received more Neutral votes / abstains (that is, fewer Supports or Opposes). The mild trend from top-left to bottom-right in 2008 suggests a tendency by voters not to cast a vote for candidates other than those they preferred. No such trend was evident in 2009, suggesting that more voters deliberately opposed or consciously left the radio button on default for the candidates they did not prefer.

It is striking that a line can cleanly divide the candidates of two ArbCom elections in this respect. Several possible reasons for the shift have been put forward. Short Brigade Harvester Boris that "we know what happened, but we can't say why it happened". Among the possibilities, he said, are that automated secret balloting lets people vote their conscience without fear or favour, that the simple radio-button format makes it easier for people to vote for/against, and that more candidates stood in 2008, so people didn't have time to look into all of them. John Vandenberg has raised the effects of voter fatigue from visiting so many pages, and the inevitable edit conflicts that arise from scroll-and-type voting, as possible reasons for the decrease in neutral/abstain voting. Coren that "while why is an interesting question, the effect itself is notable and (IMO) desirable: the number of actual opinions expressed [for or against] has almost doubled."

The second shift occurred in the level of Oppose votes. Figure 2 shows the ranking percentage (again horizontal) against the level of Oppose votes (vertical) for both years, now reversed, with 2009 above 2008. The average percentage of Oppose votes dramatically increased, from an average of 11.8% in 2008 up to 27.4% in 2009. Both elections saw pronounced tendencies to oppose the less popular candidates, a significantly greater trend in 2009. The same reasons as given above, in converse, could be proposed for this trend.

The third shift—a correlate of the other two factors—was towards a lower average ranking-percentage in 2009 (from 51.1% to 44.7%), with a narrowing of the range. This is shown in the horizontal dimension of both figures.

Feedback page
All users are invited to provide feedback on the 2009 election, including discussion of ways to improve the 2010 ArbCom election. The topics thus far include "election personnel", "the SecurePoll system", "improving instructions to voters", "voting rules", "supplementary voting", and "questions to candidates".