Maryland's Election Audits, 2016, Detail
The post-election procedure is not an audit because it does not determine that the election outcome correctly represents voter intent as reflected in the voter-verified ballots. At best, it simply verifies that the election outcome correctly represents the scan data. Further, it is not independent, because it uses the scan data provided by the voting system to audit the very same voting system. It does not independently confirm that the scan data represent the ballots.
Additionally, the plans appear likely to violate budget reporting requirements, which require the state to carry out an audit (one would be hard-pressed to call this an audit), to say why hand and eye inspections of paper ballots were not necessary, to report on the risk-level of the audit, and to describe the process by which the public could provide comments on the audit before it was carried out, and how the public could observe the audit. (Observation is meaningless when unverified electronic data is processed by software that provides no mechanisms for verifying its computations or input or output data). Finally, if Maryland believes the scans do represent voter intent and intends to base its audit upon the scans, the scan data should go through the procedures of secure custody of data---which include published cryptographic commitments or digitally-signed data---but the proposed procedure does not mention any. I have written to committee chairs in the Maryland Legislature to inform them of our concerns and our offer. From comments made by the administrators in the 28 October meeting, I gathered that they were concerned about the cost of an audit and the manpower planning that could be required.
For a ballot polling audit, where ballots are chosen at random from all relevant collections, and a risk-limit of 5% (if the contest outcome is wrong, the chance that the audit will conclude the outcome is right is at most 5%), assuming the counts are all correct, Maryland will need to sample, on average the following numbers of ballots (we used counts published as of about 7 pm, 10 November 2016; the numbers have likely changed slightly as the counts have been modified due to inclusion of early voting and absentee votes):
These numbers could increase once we take over and under votes into account, and once all absentee ballots are in. But they give a reasonable estimate of the order of magnitude of the effort required. Our assistance in carrying out this audit will be at no cost to the state.