`Procedural Improvements` of data collection methods should never come during the gathering phase
I came across this article regarding the ballot audit in Arizona over the 2020 election.
Summary: An Arizona democrat volunteered to be an audit observer on Monday April 26 and Tuesday April 27 2021. She raised concerns about what she saw at the audit, saying she heard monitors say “speed is more important than accuracy”. She opines that the process should be the same from beginning to end. The audit must be completed by May 14, but is nowhere near completing the audit in time at the current rate - it would need to triple its current speed. A republican audit liaison responded to the concerns addressed. In particularly, the liaison said that procedural improvements should be made every day while the audit is ongoing.
Analysis: From what I understand, this “audit” is really more of a recount, the main (only?) difference being that the official results of the 2020 election will remain as they are regardless of the findings. The counting of ballots can be regarded as data collection, and as such, should follow certain rules established and tested over centuries of scientific development.
We can view an election as a type of data gathering endeavor that follows the scientific method. The scientific method involves an initial problem or question, a hypothesis, a test of the hypothesis and research gathering, concluding whether the hypothesis was supported or not, and reporting on the findings.
An election follows that methodology pretty closely. First, there is a question: who will receive more votes? Since the possible outcomes or answers to that question are limited to a few possibilities, and they can all be tested via the same experiment (the counting of the ballots), the hypothesis stage isn’t necessary. The experiment takes place in two phases: the vote and the count. When a conclusion is reached (winner is determined), it is reported by the media.
An election is full of uncontrolled variables which affect the voting stage. People come from different walks of life, have different beliefs, different affiliations, etc… The presence of these variables is essentially the reason for democracy itself. If society was a monolith, we could easily be represented by the few or the one. An election needn’t account or control the variables that might affect who someone would vote for.
The thing about elections is that the will of the electorate is supposed to effect the outcome. That’s the point. Therefore, the process must treat each vote and each voter equally. The process is the control.
So, Arizona Republican liaison Ken Bennett making this remark about the audit in process is concerning:
Yes, procedures can (and should) change every day as improvements are identified. Again, I have personally been in team leader meetings late in the evening when team leaders share ideas to make improvements. Most are small incremental changes but some are significant (like removing the video monitors and just looking at the ballots).
Any change made during an experiment can affect the outcome of that experiment. “Procedural improvements” are made to increase speed, accuracy, and/or efficiency. Such changes can affect how many errors are made in the process – how many ballots are counted correctly.
This may not matter so much if the ballots were sufficiently randomized, which is unlikely the case. If they weren’t, then any procedural changes mid-count would mean that a significant amount of ballots from one demographic (processed before the changes went into affect) would receive unequal treatment to ballots from another (processed after the changes went into affect).
In other words, the count would not be an accurate reflection of the vote, which is meant to be reflection of the will of the voters of Arizona.
Given this, a detail like this should render the audit effectively bust:
As of Tuesday, the audit had hand counted “just shy of 100,000 ballots,” according to Senate audit liaison Ken Bennett, a former Arizona secretary of state. Accepting that number at face value, the audit would have to triple its speed to finish the hand count by May 14, the day it must vacate the coliseum.
This would mean that without procedural changes, which are problematic for the reasons discussed, it would be nearly impossible to audit all the votes in the 2020 election in Arizona.
In a scientific experiment, if any inefficiencies or possible improvements are identified, a change in the process would mean starting over. That doesn’t seem to be the intent here.
references
- Resnik, Brahm. `"I Left Horrified": Arizona Audit Observer Says Ballot Counter Instructions Were for "Speed, Not Accuracy".` 12News, 29 Apr. 2021, [http://www.12news.com/article/news/politics/i-left-horrified-arizona-audit-observer-says-ballot-counter-instructions-were-for-speed-not-accuracy/75-b58c8f65-adde-4c0c-96be-166333ddb1ad].