This note is not intended to offend the fine people working on election system auditing. I hope that it will be viewed positively, and spur folks to meet its challenge.
Before getting too committed to a post election audit scheme, I suggest that we take a step back and think about what the audit is supposed to accomplish -- precisely.
Lots of folks are getting enthusiastic about potential mechanisms – the solutions. But I have not seen an adequate statement of requirements for what the election system audit is committing to accomplish. And exactly how will we know when it has succeeded, or failed?
Election system auditing is a technical/analytical/ almost mathematical problem. This is not the way I see it being approached. Sure, people are using statistics and arithmetic, but I see no postulate leading to proof.
Election system auditing is being discussed as though every ballot is like every other ballot. This is not a correct assumption. The ballot style, including its precinct, makes ballots fall into a number of non-identical sets. Something true of one ballot may be untrue for ballots of a different style/precinct.
Election system audit designers seem to not differentiate between the various methods of voting – precinct-paper, precinct-electronic, early-paper, early-electronic, absentee, provisional, and emergency. These are non-identical methods. Results from a sample in one set do not necessarily carry over to another set.
Election auditor designers seem to not differentiate between the individual units of voting equipment used to record and count ballots and votes. Just because a ballot is successfully processed on one unit does not mean that it would be processed correctly on a different unit of the same type. Each individual unit is uniquely customized for each election. No two units are identical. What is true of one unit is not necessarily true of any other unit.
And, election audit designers seem to ignore completely the processes that do not involve voting equipment, yet represent enormous opportunities for error and fraud. The voter eligibility system, the ballot issuing mechanisms, ballot production, mailing, return mail processing, signature verification, batch integrity, etc. Every one of these represents a potential threat.
In a nutshell, I hope that people wanting to be election system audit experts will first develop widespread agreement on the exact requirements for the audit and how it will be verified that the audit meets these requirements. Only then does it make sense to propose a solution.
Finally, there are data flows that make auditing possible, and data flows that do not. None of the election systems certified for use in Colorado have an auditable data flow. Consequently, all of the work being done on audit today is destined to be discarded once a more rigorous data flow model is adopted.
It is time to establish the precise requirements for election system auditing.
Colorado Voter Group
2867 Tincup Circle
Boulder, CO 80305