EVT/WOTE: When is the CT Recanvass law totally inadequate?

Editor’s Note: August 8th and 9th, we attended the EVT/WOTE (Electronic Voting Technology / Workshop On Trustworthy Elections) in San Francisco.  Over the next few days we will be highlighting several papers and talks from the conference.

We have criticized Connecticut’s recanvass law as inadequate in very close elections, where the result may depend on accurate adjudication of a handful or less of votes where voters’ intent may be critical and subject to interpretation.

In a paper delivered yesterday, An Analysis of Write-in Marks on Optical Scan Ballots, researchers analyzed 100,000 ballots for write-in votes that might not have been counted correctly. They concluded in that one sample, from one election, in one jurisdiction that 16% of write-in votes were not counted by the Diebold/ES&S/Dominion-AccuVoteOS scanners. This means, according to my math, that if a write-in candidate in a two candidate race received 46% of the vote as counted by the scanner, it is quite likely that a rigorous hand count would show that candidate as the actual winner – reversing the election night apparent result.

This is not a problem with the scanners as designed. It is a problem with voters misunderstanding the write-in process. None the less, in most states, like Connecticut, the voters’ intent should and does legally rule.

More research would be needed in more elections and on more ballot formats to determine the range of possible error in scanner counting of write-in ballots. The range might be 10% to 25% undercounted write-in votes depending on conditions. Connecticut uses those same AccuVote-OS scanners, yet a different ballot format. Our rate of undercounting could be higher or lower. In a race that was expected to be close with many write-in ballots a candidate might provide better voter education or bring out many inexperienced voters – once again, the undercount rate could vary either way.

Pending further research: To be reasonably sure that the correct candidate is officially designated as the winner in a race with a write-in candidate, it would be prudent to assume the possibility of a 25% undercount of write-in votes. Then require a recount of all write-in votes in a race where an apparently losing write-in candidate received at least 42% of the votes in a two candidate race (or 42% of the votes necessary to win in a more than two candidate race.)


Leave a Reply

You must be logged in to post a comment.