Audits of Vote-By-Mail Ballots in Whatcom County (Proposed)

Values and Policy
Background facts
Background presumptions
Recommended Whatcom Audit procedure (Summary)
Detailed Procedure
Goals of the audit activity:
  • Establish preventative conditions to deter attempts at fraud or other shenanigans.

  • Prove the validity of the published results. Catch any differences between machine-count and hand-count, so they may be investigated and resolved before the election is certified.

  • Enhance public confidence in the election system.
Values and Policy We insist on:
  • always having a voter-verifiable paper ballot for all votes cast, and
  • always using those papers as a part of the routine of every election, for a truly independent double-check on the results of whatever method was used to count the votes, and
  • always doing that double-check (commonly called an "audit") in a way that demonstrates the validity of even a computer count of votes by also counting the votes on paper and comparing those totals, and
  • by the existence and public announcement of this quality-control mechanism, be a deterrent to the interference of any actions intended to mess with votes or vote totals, and
  • discover accidental errors in the vote-counting system, and
  • bring voter trust in the validity of the system of voting and vote-counting to a high level of confidence.
Background facts:
  • Whatcom County has 119 precincts.
  • 4 of those precincts are very small (less than 100 voters), thus are not representative of a significant percentage of the total votes cast in any election.
  • 3 others have less than 200 voters - and the largest precinct had 1,584 (1.70%) as at Dec 2004.
  • Voting at the polls (even with VBM, there will be sites for the disabled) begins at 7:AM and ends at 8:PM on election day.
  • State law allows elections offices to prepare (but not count) ballots during the ten days before election day, and it allows counting to begin on election day, but no results of totals may be announced until the close of the polls (8:PM).
  • Custom has the first unofficial totals calculated and announced about one hour after the close of the polls (thus about 9:PM, election night.) The Whatcom County Auditors' office distributes those preliminary results with a printout handed to the public in the lobby of the courthouse, and on their website.
  • The smallest set of vote totals announced for principal races is the totals by precinct (averaging 800 registered voters and often up to 600 ballots actually cast, with a 75% turnout)
  • Computer program codes regularly contain conditional instructions, such that processes like totalling will be done differently in response to specific conditions found within the data. ("IF this special case is found, THEN count with rule 1, ELSE count with rule 2".) Such techniques could produce totals which do not coincide with the true totals of the data input - i.e.: totals can be intentionally corrupted.
Background presumptions:
  • It may be costly and time consuming to double-check (to audit) 100% of the ballots, though that clearly can be done if and when it is needed.
  • Checking a sample of the ballots, within suitable time-frames, is sufficient as a deterrent to intentional tampering and a demonstration of accuracy of the machine count.
  • The smallest set of vote totals which gets officially announced, thus is publicly available for comparison, is a whole precinct.
  • If a malicious act of changing vote totals were planned by someone with access to the system, knowing -- ahead of time -- which precincts would be used for audits, would allow them to direct the system to report those totals correctly, while allowing the totals in other precincts to be distorted.
  • Similarly, if a malicious act of changing vote totals were planned by someone with access to the system, knowing the preliminary results after the first tallies were reported could allow direction to influence the selection of the precincts to be used for audits (and reporting their totals correctly), while allowing the totals to be distorted in other precincts which would not be audited.
Recommended Whatcom Audit procedure: See Detailed Procedure:

Simplified steps:
Randomly select some precincts and some races to be audited. Do the selection in a fair manner that minimizes the opportunity for interference.
An allowable margin of error will be pre-established.
Hand-count the paper ballots sometime after many of them have been returned, yet before the certification.
The Auditor and citizen observers will compare the totals and differences by machine and hand count, and proceed to reach and announce a confident determination of the true outcome.

See Detailed Procedure:

(page presentation edited August 2006)

back to NoLeakyBuckets homepage