|
December 11,2024
Georgia County Election Officials,
This morning at their meeting, the State Election Board will be discussing the recent OCR Ballot Image “Audit” exercise the SOS required that you undertake after the election. Last night, we sent the following email to the Board urging them to correct the record during their discussion today so that the public and press are not misled by Secretary Raffensperger’s misinformation concerning the “audit” exercise.
While we support the valuable OCR image analysis and exercise, it is important for county election officials, voters, and candidates to have actual facts about what that post-election exercise did and did not accomplish. We hope that the Board will ensure that the information disseminated to the public today is more accurate than the Secretary’s politically motivated press statement. Your voters deserve facts, not spin from the SOS.
Coalition for Good Governance Statement on the November 2024
OCR Ballot Image Analysis
Given the SEB’s planned review of the OCR Image “Audit” at the December 11 meeting, we ask that the SEB take this important opportunity to correct the misinformation Secretary Raffensperger has disseminated concerning the OCR image exercise. It is important that the public not continue to be misled about the attributes and findings of the image analysis. The benefits and limitations should be objectively and factually stated, which will require contradicting Secretary Raffensperger’s baseless claims.
We offer the following comments on the OCR image analysis and Secretary Raffensperger’s misleading statements about the so-called “audit.” We refer to it as the image “analysis,” because an “audit” of the accuracy of an election outcome is a very different exercise with different goals and meaning than the limited scope exercise conducted.
--The image analysis conducted was a worthwhile exercise, provided valuable information, and detected certain types of discrepancies, which were then corrected. Such analysis should be continued in future election, but caution is necessary to avoid overstating the meaning of the results of the image analysis in public communications.
--Analyzing digital ballot images electronically is a far cry from auditing the outcome of an election which requires manually reviewing appropriate samples of original trustworthy ballots. Digital images are not inherently trustworthy records and the election outcome cannot be verified by analyzing digital images, especially when that analysis is done electronically.
--The SOS press release (below) misleadingly states that the OCR analysis confirmed the “accuracy” of the counties’ reported results, even falsely claiming “zero differences.” That is not true, or even possible, for multiple reasons:
-As reported by the vendor, Enhanced Voting, multiple counties had to make corrections after the initial certification, after the analysis found errors requiring recertification. The review triggered the corrections, confirming that the results were initially inaccurate and had substantially more than the “zero differences,” Raffensperger untruthfully claimed. Further, hundreds of vote “adjustments” and reconciliations were done to get the OCR vote count and the reported vote count to match. The public had no opportunity to observe those in most cases.
-When ballots are adjudicated inaccurately resulting in erroneous tallies, the OCR review will not necessarily detect it. (For example, we observed ballot images in this election with write-in votes for qualified candidates whose names were also listed on the ballot, but those valid write-in votes were overlooked by both the original count and the OCR analysis, resulting in an inaccurate count.)
-Ballots which are double scanned and double counted, or failed to get scanned are not detected or corrected in this type of analysis. (Double scanning ballots was a significant problem in 2020, causing thousands of votes to be inaccurately tabulated. It is expected that some double counted ballots may be present, although in much smaller quantities in this election.)
-The analysis does not ensure that the proper ballots were all counted once and only once. For example, in reviewing the HD105 recount (with a 80 vote margin), we noted 15 more mail ballots counted than mail ballot voters whose ballots were accepted.
--The analysis of the BMD ballots or their images cannot provide any assurance that the outcome of the election reflects the voters’ will, given that the computer-generated ballots are not trustworthy records of the voters’ selection. The compromised voting system is extremely vulnerable to hacking and mis-programming that can cause the printed text to reflect different choices than made by the voters, and tend not to be noticed by the voters. No analysis or “audit” of any kind can detect and correct such discrepancies for BMD ballots, which are not trustworthy source documents for an audit.
--Secretary Raffensperger misleadingly states in the press release below that the OCR analysis showed that the ballots were counted “securely.” There is no meaningful assessment of security that can be made by this analysis of digital images. This claim is baseless and untrue, and should be corrected by the SEB.
--Raffensperger stated that the image analysis shows that the system “can be trusted.” The analysis certainly shows no such thing. The top voting system cybersecurity experts have long warned about the extreme and alarming dangers of the compromised voting system, vulnerable to undetectable hacks and electronic subversion. CISA verified the insecure nature of the system in 2022. The leading experts in the nation agree that Georgia’s system and software should not be trusted. This analysis did nothing to change that.
--Raffensperger falsely states with no basis whatsoever that this exercise illustrates that computer generated ballots are “more accurate” than hand marked ballot. The analysis had absolutely no test at all of the accuracy of BMD ballots in recording the voter selections. His claim has no credibility. The scientific literature is clear that BMD voters make substantial errors and very often do not detect and correct them. BMD ballots conceal voter error, and errors go undetected. Hand marked ballots reveal voter error, and officials can almost always determine voter intent to accurately count the vote.
--No one can say whether the QR code and the tabulations of them were “accurate” reflections of the voters’ touchscreen votes, but the SOS Press Release states, again with no basis, that “the QR codes were 100% correct in all cases.”
“Magical thinking” about BMDs seems to have overtaken Georgia’s election officials about the fitness of the BMDs to be used in public elections, and what can be learned from an OCR analysis of the ballot images. We urge the SEB to recognize the limitations of the electronic image analysis that was conducted and make a point at the upcoming meeting to thoroughly correct the numerous inaccurate claims made by Secretary Raffensperger about the exercise. We encourage the SEB to support the continuation of OCR image reviews by promulgating rules that maximize the value of the exercise while realistically recognizing the limitations.
The body of the SOS press release is included below for reference.
Marilyn Marks
Executive Director, Coalition for Good Governance
Marilyn@uscgg.org
(SOS Press Release) November 25th, 2024
Atlanta - Georgia Secretary of State Brad Raffensperger announced the results of the ballot image audit conducted in coordination with Enhanced Voting and local officials in each of Georgia’s 159 counties. The audit examined 5,297,262 ballot images from thousands of Dominion Voting System tabulators deployed during the 2024 General Election. The audit utilized the Enhanced Audit software from Enhanced Voting to examine the ballot images and provide county election officials with the ability to review ballots and potential discrepancies. This image audit of the 2024 General was a major step forward in auditing technology and practice.
“The ballot image audit shows again that the votes in Georgia were counted accurately, securely, and quickly,” said Secretary Raffensperger. [CGG: Inaccurate. See above.] “I’m proud of the hard work of the elections staff in my office and the county election workers across Georgia.”
There were zero differences [CGG: Inaccurate. See above.] found between the tabulation based on the human-readable text and the tabulation based on the QR codes generated by the Dominion ImageCast X ballot marking devices. In every instance, the human-readable text on the paper ballots matched the votes represented in the QR code on that ballot.
There were discrepancies found on 87 ballots out of the 5,297,262 cast. That represents 0.001642% of the ballots. All 87 were due to human reviewers attempting to discern voter intent. [CGG: These discrepancies do not include the test ballots counted and inaccurate tabulations because of misprogramming, both of which were reportedly corrected after the image analysis detected them.] One ballot produced by a ballot marking device had a write in in a local contest reviewed and adjudicated incorrectly to a candidate on the ballot. The remaining 86 discrepancies were on hand-marked paper ballots. There were 271,399 hand-marked ballots cast, which yields a 0.0317% discrepancy rate. This illustrates that ballots produced by ballot marking devices are much more accurate than hand-marked paper ballots. [CGG: inaccurate.]
The audit examined every race in every county in Georgia. There were 1,955 total contests in Georgia in the 2024 General Election. That includes 295 federal or state contests along with 1,660 local races. This audit examined each ballot cast in all of these races. The QR codes were 100% correct in all cases. [CGG: inaccurate]
“The image audit shows that our system works accurately and can be trusted,” Raffensperger added.
| 
