Secretary Bysiewicz Orders Expanded Post-Election Audit

Update: 8/20
Checking with Dr. Shvartsman about the Secretary’s “98% or 99% probability”, he said:

Indeed, 10% sample from the population of about 800, counting all CT districts, gives about the same statistical result as 33% sample from the population of about 250, counting only the districts involved in the August primary…there is no inference of any kind to be made about adequacy/inadequacy of the 10% sample. The recommendation was based purely on the basis of the population size being reduced from about 800 to about 250.

In general I agree with Dr. Shvartsman. If we were auditing one race over any reasonable number of districts, then what is important is the sample size so 10% x 800 = 80 and 33% x 250 = 83.

For instance, if we want to do a poll of Connecticut and New York voters, if we sample 1000 voters in each state, the resulting confidence is about the same, even though the number of voters in each state is very different.

And given the nature of Connecticut’s audit law and the interaction of the 10% district selection followed by the random selection of races, it is a wholly different calculation.

Further what the Secretary of the State seemed to be claiming had to do with a proof that the machines were working properly, which is a whole different situation – hard to even define given the nature of the audit.

*************
Original Post:
Secretary more than triples legislature’s 10% audit mandate for primary, 82 districts in 33 towns to be audited.

Along with several advocates, I participated in the post-election audit drawing this morning. Secretary of the State, Susan Bysiewicz indicated they were “advised by Dr. Shvartsman of UConn” with such a small number of districts that this larger sample would provide “98% or 99% probability” that the machines functioned properly. She also reiterated that the November 2007 post-election audits showed that the machines performed “extremely accurately and securely” the only errors were those of voters filling out ballots incorrectly – We dispute that.

In general I am pleased to learn that the Secretary of the State sees that the current audit law is inadequate and that statistics should play a role in determining the level of post-election audits. On the other hand, I would spend additional and existing post-election auditing/recounting resources in ways that would provide higher levels of confidence more efficiently and effectively. These mechanisms as articulated in our petition include:

• Variable audits by race, based on the number of districts and originally reported margin (i.e. much lower audit levels for races such as the biggest one in this election, the 4th congressional district primary with a margin over 70%)
• Dropping the exemption of districts with recounted races from the audits
• Dropping the exemption of centrally counted absentee ballots from the audits
• Re-instituting hand-counts for close elections – the Secretary supported them for two elections, but now has gone back to machine recounting – one race in this election in New Britain was won by three votes, such races should be recounted by hand (as of this point we have not verified if this election was recounted by hand or was recounted by machine)
• Dropping the exemption from the audit of counting ballots that were originally counted by hand – we support auditing all ballots – if the November 2007 audits proved anything, they proved election officials do not have confidence in their ability to count ballots by hand accurately without independent verification;
• And with a chain-of-custody we can trust, registrars can follow, and that is enforceable.

The Secretary also indicated that this audit would prove that the machines were ready for the November election. Once again, CTVotersCount respectfully disagrees: The memory cards are programmed before each election so one post-election audit is limited in its ability to provide insight into the integrity of another election; This may help verify that the machines generally perform well and may not have been compromised by the mandatory vendor (LHS) maintenance performed on the machines this summer; However, as we have said and UConn reports have shown the machines can be programmed to change results based on all sorts of criteria. This makes real the Computer Science reality that any test cannot be conclusive.

Speaking of UConn and the Secretary of the State’s office, we continue to await the promised memory cards test reports from the February Primary and the legally mandated post-election audit reports from that same primary. So far we must go with the Coalition reports for that primary which indicate that due to procedural lapses the value of those audits remain questionable.

We half agree with the Secretary’s statement today:

“Auditing election results isn’t just a good idea, it’s absolutely essential in order to guarantee the integrity of our elections,” said Secretary Bysiewicz. “As Connecticut prepares for perhaps its highest turnout election in a generation, it is important voters have faith that their vote will be recorded accurately and that’s why the independent audits are so vital.”

Yet, we would prefer to have “confidence” rather than “faith” that everyone’s vote will be recorded accurately.

List of towns and districts: