It Is Not Whether Or Not To Audit Elections, But How,
As Explained in Short Paper and Spreadsheet from National Election Data Archive
Salt Lake City, UT - January 17, 2007
The National Election Data Archive (NEDA) has released a short paper explaining a new formula developed by Ronald Rivest of MIT to estimate the minimum audit amounts that are mathematically sufficient to detect vote count errors that could seat wrong candidates. NEDA's paper and an easy-to-use spreadsheet to allow any layman to calculate how many vote counts to audit for a particular election contest, can be found at ElectionArchive.org.
There has been a nationwide debate as to whether or not to audit election results to verify machine counts by manually counting paper ballot records. The Federal Deposit Insurance Corporation (FDIC) insures deposits in banks because banks are subjected to certifiable audits. The National Election Data Archive (NEDA) and other election integrity activists want elections, one of the very foundations of our way of life, to be held to a standard as high as the banking industry.
The debate has now shifted to whether or not election audits should be statistically meaningful, mathematically sufficient, and transparent or merely use small fixed rate samples insufficient to detect cases when machine vote miscounts have reversed election outcomes.
As much as some people believe that an audit of a small fixed percentage of vote counts is adequate, fixed rate audits are not always adequate to detect vote miscounts that could put the wrong candidate in office.
If the purpose of an election is to carry out the will of the people, then the purpose of an audit is to make sure that election outcomes accurately reflect that will. Anything less than a manual audit that is sufficient to detect vote miscount that could put the wrong persons into office is just a "Trust us", pseudo-audit that gives the public a false sense of security.
An audit that is not transparent, publicly verifiable, and statistically meaningful just wastes time, effort and money because it does not ensure that election outcomes are correct.
A chart at the bottom of the "HowManyToAudit" spreadsheet demonstrates the shortcomings of using fixed percentage audits to uncover vote count errors, i.e. In an average US House race with a 1% margin between candidates and 440 precinct counts, a 2% audit may only have 27% chance of uncovering vote count error, whereas a 20% audit may have an 97% chance of uncovering vote count error.
Other requirements to ensure the integrity of election outcomes are included in a set of 14 Recommendations for Ensuring the Integrity of Elections by experts in election integrity. Sufficient audits and public oversight are necessary to deter wholesale electronic fraud and errors.
For the public to have oversight over election integrity, the public needs to be allowed to fully observe all audit procedures, including the random selection and manual vote counts, and before the audit begins, a public report of all vote types and votes counted on each machine needs to be publicly released. Audits are hand-counts of all paper ballots associated with randomly selected machine counts.
Today, there is no state in America which audits sufficient vote counts using adequate audit procedures to ensure the integrity of all its election outcomes.