The Arizona Printer Problem
I am pretty technical, and I have performed troubleshooting for some time in multiple industries and with multiple technologies. One of my first rules for troubleshooting any problem is scope. What is the scope of the problem?
To really understand the scope, you must have clarity on the problem. Maricopa County has now admitted there was an issue or issues on election day 2022, and they are undertaking a root cause analysis. The question I did not hear is: a root cause analysis of what problem(s)?
Given the admission of issues, I will frame my problem statement this way:
Were enough ballots printed and not counted at the voting center to impact the results of the election, i.e. is the door #3 count the representation of the impact of the issues? Along with that, did the county accurately report the scope of the problem to the public and to the court? Did the problem impact one group more than others?
Proposed solution: This is a data-driven system with several different databases (an educated guess). First, query the “check-in and print system.” I would pull at least the following information: Voter identifier, party affiliation, check-in date and time, ballot identifier, polling location, terminal ID where the voter checked in.
That will determine the number of people who physically visited a polling site on election day and checked in, and were given a ballot.
Second, using a standard SQL query, join with the table of ballots scanned on election day. There are several ways this could be done, and any database expert should be able to work with the tables and extract this information. With access to the system and a court-appointed expert monitoring me, I would estimate I could pull the data in less than eight hours. Two to four hours if data dictionaries are supplied by the vendor.
I would also generate a list of ballots not accepted or counted at the polling location, to be used later.
Now, we have the total number of check-ins and the total number of ballots scanned by location on election day. What is that difference? Keep in mind that these election machines were certified, but to maintain certification, the systems should have an error rate of less than one percent. I understand this is a separate issue, but if they were certified incorrectly, that may be an issue to be debated later.
If these numbers don't match, this is closer to the true scope of the problems, not the false door #3 counts. Door #3 only represents a portion of the total population impacted. It represents the people who put the ballot into door #3. I suspect that is why they keep pointing the court and media to the door #3 estimates because the number of impacted voters is higher.
By pulling party affiliation and location, you can now determine if the numbers are significant, how many actual locations were impacted, and if one party was more impacted than the other. I would drop that into Excel and summarize the results by location and party. I would potentially summarize by check-in station to determine if there was a pattern there.
This will also validate whether the door #3 numbers are the full population of issues as presented by Maricopa County or not. To state it differently, this will determine the full scope of the problem statement of how many ballots were printed and not scanned onsite. This should be the problem statement that we are trying to resolve.
The next problem statement that I would look at is:
Of the votes that were not counted at the polling center, what were the number of votes scanned and counted at the main tabulation center? Again, this will validate the assertion that all ballots were counted.
Taking the list of ballot IDs not counted at the polling location and running that list against the ballots accepted and counted at the main tabulation center would then validate the number of ballots that had an issue at the polling place being counted.
The final audit check I would perform is to take some samples from this data set and run them through the machine to confirm the proper tabulation of the ballots. I.e. hand count 200-300 ballots that you run through the machines and confirm the results. Optical readers are set pretty precisely, and reading a ballot not printed perfectly could directly impact the ability to read that ballot.
I think it is possible, with the right technical expertise, to properly scope the problem and answer the question -- were enough voters impacted by the problem to change the election? And were all the ballots cast properly counted?
I write this in part because Maricopa County and the Secretary of State seem determined to swear in Katie Hobbs. Maybe she won the election outright. But after admitting there was a problem, the courts now owe the good people of Arizona a thorough vetting of the problems, the troubleshooting, and the final counts. And, after misleading the court several times (for example, only 20% of the locations were impacted was declared in court on election day), there were no page size issues, etc. I think the court should have some skepticism of the county’s claims, and the court should appoint an expert to audit the various steps in the process to determine the accuracy of the declarations made to the court by Maricopa County officials. If declarations were made misleading the court, I think sanctions should be considered.
If the data has not been tampered with, this process would determine how many locations were really impacted, how many check-in stations had issues, and if there were any ballots printed and not counted at any time. Good luck, Arizona.
Maker S. Mark (a pseudonym) is a patriot who can understand and explain advanced math and science, and is worried about the state of the nation and how to solve the problems we face.