Evidence-Based Elections

We favor “Evidence Based Elections”.  We recently reread this 2012 paper by Phil Stark and David Wagner,  Evidence-Based Elections <read>

It covers at a high level the requirements to provide the public and losing candidates the evidence necessary to convince that its very likely the candidate favored by the voters actually was declared the winner of an election (or determining, if possible, the winner).

  • Paper ballots (To date there is no other viable voter-verified record).
  • Software Independent Voting Systems – the whole system, computer, human etc. can produce an accurate result (independently) even if the computer and software systems are in error.
  • Compliance Audits – that the election was conducted as intended. e.g. we can trust the paper ballots and the check-in records.
  • Risk-Limiting Audits – that demonstrate that there is a certain chance that if a contest was wrongly decided, the audit would have detected that.  e.g. 90% or 95%.
    (A 95% detection risk does not mean that there is a 5% chance that the election was wrongly decided. Only that if there was error or fraud 19 times out of 20 if would be detected e.g. if there was a 95% chance a person would be caught each time they used a cell phone while driving, few would risk it.)
  • The overall election and canvass process should correct its own errors.

Finally, the authors point to the limitations of certification and testing of election equipment and the advantages of easing the constraints of setting unrealistic expectations for certification requirements.

Sadly, no state has full risk-limiting audits.  Only about half have audits at all. Few have compliance audits.  About half have close-vote recounts, which provide self-correction when the initially reported results are close.

Compared to all the states in the Union, Connecticut would rank slightly above average, yet far from approaching credible evidence-based elections. We have paper ballots, inadequate post-election audits, close-vote recanvasses, no compliance audits, and atrociously weak ballot security.  This is a case where rating/ranking should be the result of multiplying the factors, rather than adding them:

Paper Ballots(1)  x  Post-Election Audits(0.3)  x  Self-Correcting(0.4)  x  Compliance(0) = 0




Leave a Reply

You must be logged in to post a comment.