How Easy To Audit Elections With 95% Confidence?
How Easy Is It To Audit Elections With 95% Confidence-level?
A new analysis using Utah's 2004 General Election results, finds that obtaining 95% confidence-level in election outcomes only requires auditing a reasonable 5.5% of all precinct races overall.
The National Election Data Archive has created a Powerpoint presentation showing simply "How to Determine Initial Sample Size for 95% Confidence-Level Vote Count Audits"
Mandatory Post-Election Vote Count Audits http://electionarchive.org/ucvAnalysis/US/paper-audits/UT/MandatoryVoteCountAudits.ppt
and a smaller pdf version of the powerpoint presentation: http://electionarchive.org/ucvAnalysis/US/paper-audits/UT/VoteCountAudits.pdf
Counting overall 5.5% of Utah's precinct-race counts using this method achieves a higher confidence-level (95%) at lower cost than any flat manual election audit such as 10% audit Connecticut uses.
Utah's overall audit rates for the 2004 general election are:
For all federal and state races audited - 5.4% of precincts
* Federal & Statewide Races total precincts audited - average 3% of precincts
* State Senate Races total precincts audited - average 11.2% of precincts
* State House Races total precincts audited - average 30.9% of precincts
The minimum required audit rate for a race is 1.5% and maximum audit rate is 100% for one close Utah State House race. Audit amounts will vary state by state depending on the number of close races and how many vote counts there are in each race.
Typically a larger overall percentage of precincts must be audited, but not a larger "amount" of precincts, for races with fewer total number of precincts. The last slide in NEDA's powerpoint presentation explains why a flat percentage audit of precincts (or other vote counts) - the approach adopted by most states currently - is mathematically incorrect.
A general rule of thumb for that can be applied without using any fancy mathematics, was suggested by M.I.T. Professor Ronald Rivest: An estimate the audit sample size for any race can be obtained by dividing one (1) by the margin (in percentage of ballots) between the winning candidate with the least number of votes and the runnerup (the losing candidate with the most votes). This easy calculation (1/margin%) is not precisely accurate but provides a confidence-level of somewhere between 70% to 100%. Eg. if the margin is 1%, then manually audit 1/0.01 = 100 vote counts; and if the margin is 50%, then manually audit 1/0.50 = 2 vote counts. More precise methods to determine vote count audit sample sizes are described in the National Election Data Archive's powerpoint presentation.
The powerpoint presentation also includes a description of a new method by Aslam, Popa, and Rivest for weighting random selection of precincts by the amount of possible error that the precinct could contribute to the margin between the winning and losing candidates.
NEDA is also releasing a spreadsheet to make it easy to see how to calculate vote count audit sample sizes using unofficial election results: http://electionarchive.org/ucvAnalysis/US/paper-audits/UT/GeneralElection2004-Audits.xls
MODEL LEGISLATIVE PROPOSAL FOR 95% CONFIDENCE-LEVEL VOTE COUNT AUDITS:
After developing legislative language for election auditing since January 2008 in consultation with various experts, the National Election Data Archive finalized a legislative model for election auditing:
Legislative Text Requiring 95% Confidence-Level Election Audits. http://electionarchive.org/ucvAnalysis/US/paper-audits/UT/VoteCountAudit-UT.pdf
The National Election Data Archive recommends combining its bill text with a Legislative Request for PUBLIC ACCESS TO ELECTION RECORDS:
Currently no state yet subjects its election results to scientific independent election audits.
The 95% confidence-level vote count audit legislation proposed:
1. uses a flexible definition for "vote counts" so that precincts, machine counts, batch counts or even individual ballots can be the unit audited - depending on the capabilities of the voting systems (most are not auditable on the ballot level currently).
2. uses scientific definitions for "random selection" and "confidence-level" to flexibly permit any valid auditing procedure that will assure that at most 5% of incorrect election outcomes would be certified (since we expect far fewer than 100% of outcomes to be incorrect, this assures that well more than 95% of election outcomes are correct)
3. specifies procedures ensuring that audits cannot be manipulated or overcome
4. specifies specific responsibilities for auditors and election officials, and establishes an expert State "Election Audit and Recount Committee" that is appointed by expert department chairs of math, computer science, and political science departments
5. specifies publicly observable manual counts and random selections
THE TEN PRECEPTS OF ELECTION ADMINISTRATION
The Utah Voter Magazine of the League of Women Voters - Utah printed a long article I wrote about the
Ten Precepts of Election Administration, The State of Utah's Voting, Technology, and Legislative Requests for Changing Utah Election Law
With thanks to Douglas Kellner and Doug Jones.
Pages 13 to 15.
There is one small correction to the article. There are only 7 states so far which have implemented Election Day voter registration, not 22 states.
A FEDERAL RIGHT TO PUBLIC ACCESS TO ELECTION RECORDS:
Funds are urgently needed immediately by the National Election Data Archive in order to continue its legal effort to establish a federal right to public oversight over the electoral process.
The Freedom of Information Act (FOIA) does not provide public access to election records, even in federal elections because FOIA only applies to records owned by the federal government. Election records are owned by local state, county, and townships. A public right to access election records for all federal elections must be established if there is to be public oversight over the electoral process.
Since June 2006, the National Election Data Archive has brought the only case to federal court that is currently suing a state in order to establish a federal right to public access to election records.
Even with scientific vote count audits, if there is no public access to view election records, then the public cannot verify the integrity of elections. It took the National Election Data Archive 1.5 years to bring its public access to election records case into federal court. We have spent $5,000 on the case so far, but are virtually out of funds and may need as much as $20,000 more, or alternatively pro-bono or contingency attorneys to finish the case.
PLEASE help us financially or we will have to drop this crucially important legal case within a few weeks.
There is no other case currently in federal court that is poised to establish a nation-wide right to public access to election records (and hence public oversight over elections). There is also no federal legislation proposed currently in Congress to establish a federal right to public access to election records.
If we have to drop this federal court case now due to lack of funds and legal help, it could take another two years to bring another similar legal case to this point. This is crucial and without your help our federal legal suit to establish a federal right to election oversight and access to records will be dropped. Here is our current legal filing with the court:
Our legal case is ideally suited to establish a federal right to public access to election records because:
Utah current election statute is possibly the most secretive with respect to public access to election records of any State currently. Clearly Utah law violates of federal statute such as the National Voter Registration Act which requires public access to voter registration records.
Utah current election statute is clearly unworkable because it is being violated during every election by Utah's own election officials in order to perform their jobs of tallying votes.
Prior federal case law has established that the right to vote includes the right to have that vote accurately counted, and this right to have votes accurately counted cannot be verifiably assured without the right to "know" that votes are accurately counted, and this knowledge cannot be established without a public right to access election records.
To establish a federal right to public access to election records only requires winning that right in one federal case in one state. The right to public oversight over elections via public access to election records can be won here in Utah for the entire nation, but only if we have funding and legal resources to continue the case that we have spent almost two years getting to federal court.
Here is how you can donate to help fight for a federal right to public oversight over elections, and finish the mountain of work that is still required to make it possible for U.S. elections to become verifiably transparently accurate.
Please donate now. We need to raise at least a few thousand dollars immediately, and perhaps as much as $20,000 to cover all costs for this federal court case to establish a public right to access election records for the entire nation.
In Pima County, AZ, the Arizona court recently decided that the plaintiffs had a right to obtain and examine copies of the Diebold/Premier's GEMS server vote count database. Following the 2004 election, an AK court decided that the Alaska Democratic Party had a right to obtain copies of the Diebold GEMS server vote count database. Unfortunately neither of these cases establishes a nation-wide right for public access to similar election records. We need funding NOW to continue the only federal court case that is positioned to achieve this right for the entire nation.
Without a public right to access election records, there can be no public oversight over elections. It is urgent that we obtain funds now or we will be forced to drop this particular effort.
There have been "no" responses to my prior two pleas for funding for this crucial lawsuit. Please donate now:
or mail a check to: US Count Votes P.O. Box 682556 Park City, UT 84068
If you want US Count Votes to continue our legal effort to establish a federal right to public access to election records, please donate something now. Thank you.
There is also a lot of work remaining to explain the procedures for evaluating election auditing discrepancy results in order to make valid decisions for whether to certify an election or increase the audit sample.
We need your help to get that done in a timely fashion.
VOLUNTEER POSITION OR FUND-RAISING POSITION AVAILABLE
A full-time (4 day/week) volunteer position is open in Park City, UT. Possible college credit can be obtained from your college or this could turn into a paid position for someone who can successfully organize fund-raising events.
National Election Data Archive