AUDIT METHODOLOGY AND TERMINOLOGY

Methodology

VSTOP utilizes Risk-Limiting Methodology to check for machine functionality and correctly reported election outcomes. These methods use statistical analysis designed to limit the risk that an election outcome was incorrectly tabulated and reported. Indiana deploys two different audit types using audit software developed by Dr. Philip Stark, a statistician and professor at the University of California Berkley.

  • Ballot Polling Audits – are used primarily for Indiana’s optical scan voting systems. In this audit type, votes are hand counted until there is convincing evidence that the election outcome of the audit would match the election outcome of the tabulated results. It is not meant to check if every ballot was tabulated correctly, but to holistically affirm that the voting system tabulated outcome was correct as reported. This inspection can give persons skeptical of machine tabulation confidence knowing ballots were hand counted and if the audit were to continue to a full hand count, that outcome would match the machine tabulated outcome.
  • Ballot Comparison Audits – are used primarily for Indiana’s direct-record electronic voting systems. In Indiana, every DRE system has a digital CVR (cast vote record) and VVPAT (voter-verifiable paper audit trail). VSTOP’s version of this audit type is meant to check the voting system’s functionality. The audit will hand inspect the CVR and the corresponding VVPAT record, ensuring that both records match all sampled ballots. This inspection can give persons skeptical of DRE’s due to the possibility of secret digital vote manipulation confidence that there was no post-election manipulation of digital vote records by checking them to the voter’s originally printed vote records from voting via Early Voting or Election Day.

While Indiana conducts audits a little differently, the following videos are great for visualizing post-election audits in Indiana!

Video Source: Verified Voting

Customized Video Courtesy of The Elections Group

Terminology

  • Stark Audit Tool – The Audit Software (linked above) developed by Dr. Philip Stark from the University of California Berkley Department of Statistics. It has been widely tested, accepted, and successfully implemented by a variety of State and Local Governments completing post-election audits.
  • Risk-Limit – The designated percent chance that the Audit would not detect a potentially incorrect election outcome. VSTOP’s current standard risk-limit is set at 1% for Comparison Audits and 1% to 9% for Polling Audits. The risk limit cannot be 0% because that would result in every ballot being sampled, which is a recount, not an audit.
  • Confidence Level – This is the percent the audit is confident that it would detect a potentially incorrect election outcome. Confidence Levels are variable based on the sample of ballots that is inspected. VSTOP’s current audit Confidence Level must reach 91-99% for an audit to be considered successfully completed. The Confidence Level cannot be 100% because that would result in every ballot being sampled, which is a recount, not an audit.
  • Diluted Margin – This describes the “closeness” of a contest. The smallest number of votes between candidates, divided by the total number of ballots cast in all contests.
  • Sample Size – The sample size varies based on how close the contest was. General principle is: the closer the contest, the more ballot that will need to be sampled to give us the level of confidence we are attempting to achieve.
  • Ballot Manifest/Inventory – Details the location of the ballots and how many ballots are stored in said location. This is critical for selecting of sample ballots and ensuring we know where they are located.
  • Random Seed Number – Computers are not inherently random. The seed number assists the Audit Tool in selecting the Random Sample of Ballots. VSTOP will have various individuals participating in or observing ​the audit roll a 10-sided die to generate a 20-digit number live-time. The live-time generation of the seed number ensures that no one from the county, the VSTOP Team, the voting system vendor, or members of the public can have any idea what ballots will be selected for audit.
  • Match Rate – When conducting a Ballot Comparison Audit, we are looking for a 100% match rate. The digital version of each vote record should ALWAYS match its paper counter part.
  • Rewinder – A special device developed to audit DRE VVPAT rolls. It allows the Audit Team to efficiently locate ballots selected for sample on the VVPAT rolls.
  • 6-Cut Method – Used in Ballot Polling Audits, the 6-cut method helps the Audit Team efficiently locate ballots selected for sample. Ballot storage looks very different from county to county, so this is a universal process that involves stacking a set of ballots and cutting them like a deck of cards six times and selecting the top ballot for inspection. This is done for each and every ballot inspected to ensure a random sample of ballots is selected.
  • VSTOP Tally Sheet – Used when conducting a ballot polling audit. The audit team uses this excel sheet to record voters’ selections made for each audited contest for all ballots selected for inspection.
  • VSTOP Tracking Sheet – Used when conducting a ballot comparison audit. The audit team uses this excel sheet to record whether a ballot’s CVR matches its VVPAT counterpart for all ballots selected for inspection.