NEDA Releases New Vote Count Audit Calculator
NEDA Releases New Vote Count Audit Calculator Spreadsheet
Progress occurred in "best practices for vote count audit procedures" after reading the parallel testing appendix on pp. 124-126 of the Brennan Center Report yesterday. See http://brennancenter.org
( I also glanced at the Brennan Center appendix on vote count audits and noticed that the Brennan Center Task Force recommends using the same math formula to determine number of election counts to audit that NEDA has been recommending for over a year - which is perfect.)
I discovered the following better Idea for vote count audits thanks, in part, to the Brennan Center's "parallel testing..." appendix ideas and thanks, in part, to earlier discussions which took place on the Open Voting Consortium email list.
PROCEDURE FOR DECIDING HOW MANY VOTE COUNTS TO AUDIT
The safest approach to performing vote count audits will be to calculate the minimum percentage of precinct or machine counts that would need to be corrupted to alter the outcome of any race in an election, after learning the margin between candidates in any county or township. In other words, the exact number of precincts to audit can only be safely calculated when the election margins are known.
It is important to conduct audits even if the election margins are not close and even if candidates do not request an audit or recount. This method will eliminate the need of counties/townships/parishes from conducting a 100% recount in the case of close election margins, but will mean that exactly the correct amount of vote counts are selected for auditing to detect any possible outcome-altering vote miscounts with a desired probability.
100% of the voter-verifiable paper records of a sufficient number of randomly selected electronic counts to give a high probability of detecting at least one corrupt precinct, given the minimum level of outcome-altering miscount exists, would be hand-counted.
This method does not apply to central count optical-scan voting systems that are unable to break vote counts into precinct-level counts. In that case, a statistical sampling of ballots must be compared with Election Day counts.
Whenever precinct or machine vote totals are available, an actual, rather than a statistical, audit should be performed.
The National Election Data Archive is releasing today a free spreadsheet "AuditCalculator.xls" with some of these calculations in it, See http://electionarchive.org/ucvAnalysis/US/paper-audits/
Caveat, I have not yet begun to derive the single formula to make the process easier for election officials and activists, than playing around with a spreadsheet will be. If someone else gets to this mathematical job before I do, please share your results and write them up with NEDA.
NEDA, a 501(c)(3) nonprofit, "really" could use funding to much more quickly complete a joint research paper with both mathematicians and computer scientists to finish determining and writing best practices for vote count audit procedures; and needs about $2,000/month to hire a committed programmer.
Please support NEDA's efforts by making a donation at http://electionarchive.org.
I would like to compile all the excellent vote count audit methodology that has been recently developed by various groups, including NEDA, into one paper including ideas from NEDA's earlier papers, NEDA's recently discovered and as yet unpublished discoveries, the Brennan report's audit ideas, and those from other sources. However, I've got a mountain of paperwork to finish, must prepare and give talks at several upcoming conferences and conventions, and must mail introductory funding request letters first. NEDA needs funding for, or a volunteer to act as, a NEDA executive assistant, ASAP. A part-time executive assistant could be found for $1,000/month. If every person who receives this email would donate $10/month, it would be enough to hire the desparately needed programmer and executive assistant. Patriotic persons are willing to work for very little to fight for American democracy, but we need funds.
In the meantime, this spreadsheet calculator will help activists and election officials determine how many vote counts to audit. To use the spreadsheet, just change the inputs at the top of the "#precincts" worksheet (including the number of total machine/precinct counts in your county/township/parish) until you obtain the probability you want.
http://electionarchive.org/ucvAnalysis/US/paper-audits/ in the updated "AuditCalculator.xls" spreadsheet.
You can determine the number of vote counts (machine or precinct) to audit in your own county to ensure with high probability, detecting any out-come altering vote miscounts.
The third worksheet in the spreadsheet, "#precincts", is the one that uses the new concept, based on the assumption that at most 30% of votes would be likely to be switched from one candidate to another on any one voting machine (a reasonable assumption by the Brennan Center task force), and then calculating the minimum number of vote counts that would need to be corrupt to alter the outcome of an election in a race with a particular margin; and then using that minimum number of corrupt counts and a desired probability to determine the number of precincts/machines to audit.
The math derivations to make it into one simple formula for election officials to quickly calculate how many precincts/machines to audit can be figured out in the future - in a day or a week of collaboration.
This calculation will give a high confidence level that any miscount that could alter an election is detected in an audit.
Feel free to pass this spreadsheet's URL along to other statisticians or mathematicians who are interested in vote count audit procedures.
Note: The papers in the directory http://electionarchive.org/ucvAnalysis/US/paper-audits/ are outdated now, but the worksheet "#precincts" in the updated "AuditCalculator.xls" spreadsheet can be used (by playing with the inputs at the top) until I, or someone else, derives the exact formula to determine the exact number of precincts to audit to obtain a desired probability for detecting any outcome-altering vote miscounts in any election.
Best Regards and a heartfelt "Thank You" to all of NEDA's Supporters. Your support is much needed and much appreciated and almost (but not quite) covered the costs of my trip this prior week to the National Association of Election Director's Conference in Santa Fe, NM where I was able to raise important questions and learn new information.
Thank you for all your efforts. We all share the goal of accurate vote counts.
Kathy Dopp http://electionarchive.org National Election Data Archive Dedicated to Accurately Counting Elections
Subscribe to announcements by emailing email@example.com
Please donate or volunteer at http://electionarchive.org