An audit taking place in New Hampshire has found ballots that have a fold line running through a vote bubble for a candidate can cause machines to read that fold line as a vote, causing ballots to either be counted or rejected falsely.
Philip Stark, Associate Dean of Mathematical and Physical Sciences at UC Berkeley, Harri Hursti, founder of Nordic Innovation Labs, a global technology solutions company and Mark Lindeman, the acting co-director of Verified Voting, a nationally recognized voter integrity group, are the team running the audit. They discovered folds made incorrectly on the ballots can account for a large error rate when counting votes.
“When the ballots had not been folded things tabulated perfectly,” Stark told ITN yesterday. “When the ballots had been folded, depending on the orientation, if you put them in front first, tail first, upside down or right side up, and depending on which machine you put them through you get a different rate of errors of that kind.”
“In one of the experiments I think that 72% of the votes didn’t register. It eliminated 72% of the votes for those four candidates…It’s a huge number,” Stark said.
The folds on the ballots being examined in New Hampshire were created by a machine that, due to the unprecedented demand for absentee ballots in the state for the November election, was brought in to help election workers fold ballots in preparation for mailing.
The experiments Stark and Hursti set up found a fold through the voting bubble for a candidate can result in the rejection of a ballot if another candidate in the same race had already been selected. Because the machine would read such a ballot as having been cast for two candidates in the same race (known in election parlance as an overvote) the ballot would be rejected, resulting in undervotes for certain candidates.
If no candidate was selected in a race a fold line through the bubble of a candidate could result in an erroneous extra vote for that candidate as well.
The discovery has major implications for current and future audits of ballots taking place across the country. Although importantly, Stark does not believe the errors here were deliberately caused. “You would have to know that this would happen and as far as we know no one’s ever documented that this would happen. I can’t see this being weaponized frankly.”
Comprehensive examinations of the voting machines in New Hampshire are set to begin today and the audit is scheduled to conclude this weekend.
Unlike some other audits taking place, the New Hampshire audit team is in possession of the physical ballots that were cast in November. That has led to some confusion as many social media users questioned why Stark and Hursti scanned the ballots at 300 dots per inch (dpi) instead of 600dpi, pointing out that a higher dpi makes for a higher resolution, and therefore a better quality image scan.
However Stark says the resolution of the images in New Hampshire is a moot issue as the team is in possession of the physical ballots cast. “We took the images for two reasons,” Stark explained. “The first reason we took them was to have a reference copy in case a ballot got damaged.”
“The second was in case anyone accused us of tampering with the ballots. We had a reference copy we could go to and say no this is what it really looked like.”
The scans, in other words, are not for forensic purposes. “There are other audits that are relying on the scans for forensics. We are relying on the paper which is the original.” A point Stark wanted emphasized as transparency is of paramount importance to both he and Hursti during this process.
“We’re really trying to be radically transparent about everything. Everything is livestreamed. We have a camera on everything that matters. We’re stopping and explaining what’s going on. Any artifacts we make, like the tally sheets…the logs of unusual things that people see, all of that stuff is getting posted to the web every night.”
“Everything we learn, the world knows,” he added.