Post-Election Audits – the 7.2% Audit and Other Glitches

(Note: This is an abbreviated and edited version of observations and concerns with the post-election audits recently submitted to and discussed with Lesley Mara, Deputy Secretary of the State)

It is natural for things to be learned in practice that were not anticipated in creating laws and procedures. The recent random audits are a demonstration of this. Many problems not anticipated by legislation, procedures, and yours truly. Unfortunately, the number and complexity of the problems and issues indicates that there is a need for significant changes in training, procedures, and the law.

There was no easy way to determine the dates, times, and locations of the audits other than repeatedly calling the registrar’s offices. This is especially time consuming in the case of part time registrars who don’t all return messages and don’t provide office hours on their voice mail messages. There were no other members of the public as observers in Cornwall and Hartford. The only other observer was an Assistant Registrar from an opposing party in Hartford. Voter apathy or lack of publicity? There are no notice times or publication requirements for the public audits. The only notice required is to inform the Secretary of the State’s office of the date, time, and location of an audit. For instance, a pubic audit for 9:00AM could be set at 8:45AM.

The selection of races is not required to be public. The selection of races to be audited should be public and subject to notification procedures similar to those for the audits.

The time-frame of the audits, 15 to 19 days after the election, provides too much time for the ballots and machines to be manipulated and is completely counter to the opinions of the Brennan Center. This can only be fixed in the statutes; however, the procedures can be changed to mitigate some of the difficulties: a) The random selection of districts for audit could be held much closer to day 15. b) The drawing of races should be moved much closer to the date of the audit – it could be required to be the first order of business of the same public event as the audit itself.

I was able to observe two of three audits I attempted:

Lets start with Bridgeport – three districts in Bridgeport were selected for random audit, mandated between September 26 and October 1st, yet not have been audited. The ballots remain sealed by court order. There are multiple catch-22’s here. The audit cannot be completed in the time specified by law, Bridgeport’s three districts may eventually be audited but beyond the dates for contesting the results. If a recount is ordered, then by statute the Secretary of the State must randomly choose other districts from around the state for audit – but they will not be done in the time specified by the statute – and after the time where the memory cards are required to be sealed. The best fix to all this is to change the statute to remove the loophole exempting districts with recounts from the audit and extending the date for contesting the election until after the audits are actually complete. So this leaves us with a 7.2% audit (8 districts actually audited out of 110 in the Sep 11th primary). (Note: due to a simple arithmetic error! an earlier version of this post said 6.3% counting 7 not 8 districts actually audited)

In a small town, Cornwall, the election officials performing the audit believed that the votes should be run through the scanner after being recounted by hand. They zeroed the scanner and recounted them. Despite their recognition that the procedures did not specify that and not noticing that the procedures specifically warned against doing that. Why not? Because the optical scan machines must remain sealed and constitute evidence for investigation if discrepancies or found or alleged. When well meaning people make such mistakes it is an indication that more training and supervision is necessary. This problem will not go away with experience – in a small town with an average of one audit every ten years, very few will participate in two audits, and even fewer of those will remember what happened the last time.

The newly released procedures from the Secretary of the State’s office indicate that no central count absentee ballot machines are to be audited. The statute does not seem to require them to be audited – another loophole – an open invitation to fraud.

The whole method of discounting any ballots that are in any way imperfect that were originally counted by the machine leaves a huge gap for error or fraud. And a huge gap in public perception and confidence. Public confidence in the entire process would be enhanced if the audit, counted every ballot cast in the district, and matched that to the election night reported totals. It would also be much simpler for the election officials and the public to understand. When we focus on auditing a machine we lose site of the goal of counting all the votes accurately based on the intention of the voters. This is a matter for statute.

The Secretary of the State’s procedures imply that an entire ballot must be perfect to be deemed readable correctly by the machine. Should this apply only to the race in question? It can be interpreted as only applying to the race in question. What is the intent? If it is that every race on the ballot must be prefect to expect any race to be read accurately, then what is the justification by Diebold documentation or UConn testing etc. (It is a fallacy to expect machines or people to be perfect, rational people can accept a small variance that is within an established normal range)

There was some good news. I observed the recount of two districts and one race in Hartford under the supervision of Shirley Surgeon, Registrar of Voters (D) and Chair of the Technology Committee of the Registrars Of Voters Association. This was her second audit because Hartford participated in the 2006 25-town test. The audit proceeded as intended step by step. I am also pleasantly surprised and pleased to report that the cost of the audit I observed was by my rough calculations approximately $0.05 per vote per race counted – close to the $0.04 reported elsewhere in best practices – it should be faster and more economical as more races are audited in the same district and experience is gained. (It is not relevant if the actual cost in this case was $0.04 or $0.08 it is still in a range that provides a general ballpark validation of costs)

It is easy to see that the law and the procedures are complex for voting officials, especially parttime voting officials to understand and follow. The solution is simplicity – counting all the votes for a district and comparing them to the totals reported on election night. This is also the best way to surface errors and fraud no matter what the cause.


Leave a Reply

You must be logged in to post a comment.