The whole story on the 23 Hand Marked Paper Ballots

I have a lot of folks I consider to be friends in the Secretary of State’s office, including Brad Raffensperger. I have heard my friends in that office wonder why they are consistently attacked by members of the legislature. A lot of that can be chalked up to Burt Jones and his campaign to make the secretary look, dare I say unfairly, inept. I believe the bulk of this phenomenon can be blamed on that political posturing.

But there is also a lack of trust among legislators that goes back to 2020, when the SOS’s office offered at times incomplete, inaccurate, or even conflicting information while often refusing to come to committee to answer questions. Ask me how I know. I once railed on Gabe Sterling during a committee hearing where the office told us they could not come answer our questions but then I live streamed Sterling on CNN on my iPad while we were in the meeting. And yes, video of me ranting is in the state archives.

Not all of the criticisms of the SOS’s office have been fair, and that is a blog for another time. But we call balls and strikes here, and I have to call out some really selective framing, even cherry picking, the SOS chose to employ recently in an effort to discredit hand marked paper ballots (HMPB).

You may have read the following headlines:

And so on and so on. So how did all of these media outlets reach this conclusion? To answer that we go to my very favorite headline, created by the SOS’s office itself: New Georgia Audit Reveals Every Ballot Error Came from Hand-Marked Ballots. This is the release that informed the media of the SOS’s preferred narrative that in an audit of the most recent election, the only problems were with the HMPB’s.

But that isn’t what the audit found.

Nobody reads important information anymore. They skim the headline, glance at the first sentence, ignore every relevant detail, invent the rest, and then confidently explain it to everyone else. Documents are too long. Sources are inconvenient. Context is apparently elitist. Why waste five minutes reading when you can spend twenty minutes arguing about something the article never said?

Being informed takes effort. Being loudly wrong just takes Wi-Fi.

Thank you for allowing me to get that out my system and for being a reader.

At the very end of the Secretary’s press release there is a link that directs you to the audit, only it doesn’t take you directly to the audit. It takes you to a page with links to various audits and you then have to select the right one. Which is what I did and found this glorious document produced by Enhanced Voting, the company the state employs to perform the Optical Character Recognition (OCR) audit.

FINAL Georgia Image Audit Report Download

Look, I know it is 21 pages and reporters are busy people. But if you are going to report on this subject, could you not pay someone on your staff to read the source document before you swallow the narrative whole? I mean you don’t even get to the table of contents of this report before the 23 number begins to fall apart.

On page V of the report, there is a handy little chart that shows that 7 of the 23 ballots were caused by a printing issue, and later in the report you find out all 7 of those ballots came from McDuffie County, indicating a machine failure of some sort. Another ballot is in a category called the Cast Vote Record (CVR) Export issue, another machine failure.

A rescanned ballot image overwrote another image after the CVR file had already been exported. This made the later image archive appear to contain one additional Nabilah Parkes vote and one fewer Josh McLaurin vote than the original result and CVR. That was a data export synchronization problem, not a voter-intent judgment. And yet the Secretary would have you believe this was a problem with the HMPB. It was not.

So now you are down to 15 ballots labeled as adjudication differences before you even get to the freaking table of contents, but why stop there?

So what is an adjudication difference (AD), and why does it matter?

First, remember this audit is OCR audit, meaning a computer is reading the human readable text on the ballots and giving the count. One of its benefits is that it doesn’t use the ~~Dominion~~… err… Liberty Voting software to tally the votes, and also since it is a 3rd party system, it cannot interpret the proprietary Liberty Voting QR Code. The end result is sorta like a machine version of a manual count, it only looks at the english words on the ballot images.

An AD happens when election workers and the later OCR audit disagreed about what a voter intended on a hand marked ballot. It is not necessarily a scanner counting error; it usually means two groups of people looked at an unusual HMPB and disagreed about how it should count.

I know this might get sorta wonky and won’t make a sexy headline, but you’re smart. You can do this!

A scanner may flag a ballot for adjudication because the voter:

made a checkmark instead of filling the oval,
marked partly outside the oval,
made a faint or stray mark,
selected a write-in in an unusual way,
or appeared to mark more than one choice,
the scanner was having a moment.

County officials, along with members of both parties, then review the ballot image and make a judgment: “Count this as a vote for Candidate A or B,” or “treat it as an undervote.” During the audit, a different set of reviewers looked at the same image and sometimes reached a different conclusion. The report describes these as marginal marks interpreted differently and decisions that differed from those made during the original canvass.

For example, the report shows a June ballot where the voter made a dark mark just outside the oval beside John Kennedy. The original count did not recognize it as a vote, but the auditors concluded that the mark reflected the voter’s intent.

But the most glaring omission from the SOS’s press release and the resulting headlines is that the Ballot Marking Devices also produced errors that were caught in the post election audit process and are in the freaking report for all to see.

The Enhanced Voting audit did not just cover the June primary runoff (yes, runoffs are stupid), but also included the May primary where they found that there were 16 errors created by the BMD’s. Funny how you are just now hearing about that, and from me, and only after I finally had time to read the report for my day job.

The report details how in Candler County a problem was identified to be a programming mismatch between the BMD and the scanner. After Reagan Donaldson was disqualified from the Board of Elections District 2 race, the scanner was updated to remove him, but at least one ballot-marking device was not. That left the two machines using different versions of the ballot. The BMD produced a QR code of the voter’s selections according to the old lineup, while the scanner interpreted them according to the new one. As a result, a vote for Donaldson was read as a write-in, and an E-SPLOST “Yes” vote was read as a “No.”

The audit found the problem on 12 ballots, all scanned on the same tabulator.

The error is strikingly similar to what happened in DeKalb County’s 2022 Commission District 2 race, when BMD’s and scanners were also using mismatched election definitions after a candidate withdrew. In both cases, the machines were consistently following conflicting instructions.

A single ballot in Gordon County was reported to have suffered from the same misconfiguration. For those counting at home, that is now 3 counties who have committed the same configuration error in the last 4 years, but I digress.

Gordon and Candler’s results were not changed because only 13 ballots were affected and the margins were larger, but the episode shows why Georgia cannot simply point to the accuracy of BMD ballots and declare the system flawless. The human readable text on the paper ballot is what exposed the mistake, and an independent audit caught it before certification.

The other 3 BMD errors occurred in Fulton County. On each ballot, the human readable text and the QR code agreed with each other, but the scanner attached the wrong AuditMark and cast-vote record, apparently borrowing the electronic record from a previously scanned ballot. In practical terms, the voter’s printed ballot was correct, but the electronic record associated with it was wrong, causing those ballots to be tabulated incorrectly.

The point here is not that hand-marked paper ballots are perfect; they are not. Voters make weird marks. Scanners miss things. Election workers occasionally look at the same scribble and reach different conclusions. That is why we audit elections in the first place.

The point is that this audit found problems involving HMPB’s, printers, exports, programming, scanners, ballot-marking devices, and human beings. It found an election system doing what election systems do: producing a small number of errors that had to be identified and corrected.

What it did not find was the neat little morality play presented in the Secretary of State’s press release, where the shiny machines achieved immaculate perfection while every error crawled out of a hand marked paper ballot. That narrative required ignoring the May election entirely, describing printing and data export failures as problems with HMPBs, and hoping nobody clicked through several links to read the actual report. Apparently “every ballot error came from hand-marked ballots” sounded better than “some of the errors came from hand-marked ballots, while our machines also misassigned votes, attached the wrong electronic records, and repeated a programming failure we already saw in DeKalb four years ago.”

Again, I like Brad. I have friends in that office. I also understand why they are tired of being treated like piñatas by every politician looking for a cheap headline. But you do not rebuild trust by selectively presenting your own audit. You certainly do not complain that legislators and activists distrust you while issuing press releases that omit the most inconvenient findings buried in the report you commissioned.

The audit process worked. It caught the errors. That’s the good news, and the Secretary should be bragging about that instead of choosing to turn an important election security report into a sales brochure for his preferred voting system. That may win a news cycle, particularly when nobody reads past the headline, but it does not win trust.

Published by Scot