-Aspen's historic May 5, 2009 IRV election audited as single ballots- 5/5/09 Aspen CO held an instant runoff election (IRV) for mayor and 2 council members. Interpreted contents of each ballot, scanned by True Ballot, were publicly released. Open records requests for a CD of image scans were denied. Aspen has been sued to protect records from destruction and to allow inspection of the scanned ballot files. A Court of Appeals ruling holds that unidentifiable ballots are public records.

Search this and related blogs

Sunday, December 20, 2009

How Aspen Violated Colorado Ranked Choice Voting Rules

[Mike LaBonte,  an election quality advocate from Massachusetts posted  the following summary of one portion of his work related to the Aspen election on the BlackBoxVoting.org website on January 7, and accepted the invitation to post it here as well.   --posted by M. Marks.]

Colorado state law has a section CRS 1-7-1003 allowing for ranked choice voting. It describes a form of instant runoff voting (IRV) for single seat elections, and a form of single transferable vote (STV) for multiple seat elections. It also has a clause allowing IRV for multiple seat elections with a condition:
 A local government may also conduct an election pursuant to this subsection (4) using the principles of instant runoff voting specified in subsection (3) of this section to ensure that each elector has equal voting power and that an elector's lower ranking of a candidate does not count against the candidate to whom the elector gave the highest rank.

Put another way, in an election in which candidate A is a winner, it must not be possible to have a set of ballots with candidate A ranked #1 and various other lower ranked choices, which upon changing only the lower ranked choices can make candidate A lose. Imagine that you ranked candidate A number 1 and that was your only vote in an election that candidate A wins. Now imagine you had instead ranked candidate A number one and candidate B number 2, and now candidate A loses! It can happen if the 2nd rank vote for candidate B moves B ahead of A in an early round, and A is eliminated even though A would have picked up more votes in a later round. This is one form of non-monotonic behavior. To be fair, a regular series of regular non-instant runoffs potentially has the same effect.

The method chosen by Aspen for their City Council race appears to attempt to utilize that last clause in CRS 1-7-1003 above. They did not use STV, but rather a separate IRV race for each council seat.

It can be proven with the ballot data from the Aspen May 5, 2009 election that a change in the outcome of the City Council race was possible, due to non-monotonicity. In that election Derek Johnson won seat 1 and Torre won seat 2. Briefly, if every voter who ranked Torre number 1 had also ranked Jack Johnson number 2, with no other candidates ranked, Jack Johnson would have won seat 1 and Derek Johnson would have won seat 2. Torre's victory would have changed to a loss, and his supporters would have caused it by their 2nd place rankings for Jack Johnson.

Although there is a debate about Aspen's non-release of actual ballot images, much information from that election has been made public, to their credit. Everything needed to reproduce the counting has been provided, including the "ballot string" files (interpreted from the actual ballots that we can't see) and configuration files for using the ChoicePlus Pro software used in the actual election. I have attached a zip archive with 2 subfolders. Folder "may5_original_camb" reproduces the counting using the Aspen rules, just to prove my setup replicates the original results. Folder "torre1_jjohnson2_camb" uses ballots modified as described above, in which voters who ranked Torre number 1 hypothetically defeated him by also ranking Jack Johnson number 2. It didn't happen that way. But it could have, and that is what Colorado state law seems to prohibit.

The files supporting this work are at: http://www.bbvforums.org/forums/messages/8/aspen_nonmono_irv-80803.zip

No comments: