Most of the activities termed “audits” in elections are not audits at all, but are really spot checks. An actual audit differs in these important ways:
– It is holistic, cross-checking information from a variety of source documents;
– Discrepancies trigger nonrestricted expansion of the audit;
– The entity being audited is not allowed to constrain the audit;
– A real audit involves selections based on hunches, red flags, and suspicions as well as targeting by random choice.
Why it’s important to call it a spot check not an audit: A spot check cannot prove an election was sound or that its results were accurate. Ballot spot checks can help to discover errors and may, in some cases, act as a deterrent for certain types of election fraud, but referring to spot-checks as “audits” produces false confidence and often causes election officials and the media to overstate the implications of the spot check.
HOW TO DETECT A SHAM
1. All phases must be publicly observable. Were ballots sorted or chosen before observers arrived? Was target selection fully transparent and public? Were any essential records deemed “off limits”?
2. Precinct results must be published and publicly available well before samples are selected for spot check.
3. Selection must be an unpredictable surprise. UNPREDICTABLE is not the same thing as RANDOM. Also, there should not be any unobserved time lapse between selection and spot check.
Example: Fresno County California was preselecting its spot-check precincts two months before the election. In the 2004 Ohio presidential recount, Cuyahoga County employees admitted to choosing their own precincts for a spot check, stating “our random is our random.” (Watch the film “Hacking Democracy” for footage of this scene, videotaped by a citizen observer.)
4. All ballot questions must be checkable. Random selection of one or two races is unlikely to pick the most controversial race, or the few races most at risk for fraud.
5. Selection must begin with the whole set of precincts, and public observers must have a way to confirm that the whole set was used.
Example: Several years ago in West Virginia a county election official chose a “random sample” of precincts using slips of paper in a bag. When an observer, videotaping the process, asked her to empty out the bag to show that it contained a slip of paper for every precinct, the official hurriedly left the room (holding the bag) and then returned with the bag containing all its precincts. Computer selection may seem more professional than slips in a paper bag but is actually even less transparent. Observers may have difficulty verifying that every one of the hundreds of precincts in a large jurisdiction are actually included in the spot check selection program.
5. Every ballot should have some chance of being audited. There should be no subset or class of ballots that are excluded from the audit.
Example: in San Diego County after the 2016 primary election, a large subset of vote-by-mail ballots were excluded from the spot check. Knowing that this set would be excluded, these vote-by-mail ballots could be targeted for election fraud with little risk.
6. If the audit does reveal a descrepancy it must have a mechanism to increase scrutiny, such as taking a larger sample or examining other kinds of records.
Example: In San Diego County after the 2016 primary, a number of ballots were observed with “white-out” obliterating the original votes. Regardless of whether the reason given for the white-out seems valid, this kind of discrepancy would need to be examined by expanding the study of all ballots with white-out, and by incorporating statistical analysis to determine if the white-out disproportionately affected one class of voters or one candidate.
7. A ballot spot check must be hand counted not machine counted. A machine count cannot serve as the primary way to authenticate another machine count.
8. There are four areas where election tampering can change results. A spot check or audit of one area does not cover the others. Tampering in any one of the four areas can alter the outcome.
FOUR AREAS TO AUDIT
a. Who can vote
b. Who did vote
c. Counting of the vote
d. Chain of custody
Auditing who can vote examines whether:
(1) All eligible voters would have been allowed to vote
(2) Any ineligible voters could have voted.
Auditing who did vote examines whether:
(1) Number of voters matches number of votes, and determining if turnout percentage was within normal bounds
(2) Any eligible voters were turned away
(3) Any ineligible voters cast votes
Auditing the counting of the vote examines whether:
(1) A human examination of ballots matches results reported by machine
(2) Ballot images* (pictures of ballots) match reported count
(3) Ballot images match actual ballots
- Because ballot images are an electronic record, giving public access to all ballot images can speed up verification of election results as long as the public can compare ballot images to actual ballots. (See “The Brakey Method“)
Auditing chain of custody examines whether:
(1) The votes counted included all the ballots, extra ballots, or altered ballots
(2) Whether all essential components can be shown to be the original element.
Example: Shelby County Tennessee – August 2010, a records inspection sought to determine whether poll tapes (results printed from each precinct machine) matched reported results from the central tabulator. The poll tapes were not made available until 12 days after the election. During the 12-day interval, additional voting machines were set up in an adjoining room and were observed printing poll tapes. Prior to that, poll tapes had been observed in a trash bag. Inspectors were not able to determine whether the poll tapes they were given to inspect were the originals.
* * * * *
The importance of public citizens doing field observation cannot be overstated. Success of any spot check or audit depends on getting the details right. Without vigilant public observers, official “verification” of elections can become nothing more than taxpayer-funded theatre.