User:Opabinia regalis/ACE2015 voter data

Consolidated notes on the voter-log data analysis for WP:ACE2015. (See original thread archived here.)

Vote timing

 * First three days of the election (Nov 23-26) show clear spikes in votes per hour when the mass messages went out (with surprisingly little lag time). The second spike in particular (Nov 24) is correlated with low median edit count among voters in that time period.
 * Only a very small uptick at the end of the voting period.

Voter edit counts

 * Very large increase in number of voters with relatively low edit counts.
 * Caveat: these are all 2015 edit counts (retrieved 6 Dec). However, there are not many voters in the 2015 distribution who registered in 2015 (see below), and there's no evidence of a wiki-life-cycle shift in the 2013 vs. 2014 distributions, so this is unlikely to be a strong effect.
 * Breakdown for the Nov 23-26 data:

Voter registration

 * There was widespread concern that long-dormant accounts were voting in large numbers and would offer uninformed opinions. In the end, only ~10% of voters had not edited in the past three months, which would have made them ineligible to receive the mass message under the original proposal.

Overall

 * 2674 unique voters, of whom 273 have not edited in the last three months, 128 registered in 2015, and 625 have under 500 edits.