Comparing Audtis in Different States

As part of proposed RLA legislation for NY we are listing states that are conducting RLAs, or piloting them. I’m wondering how experts categorize the New Mexico “tiered” post-election audit system. Is that categorized as a risk-limiting audit? All comments welcome. Thank you.
http://www.ncsl.org/research/elections-and-campaigns/post-election-audits635926066.aspx

1 Like

I don’t consider it a true RLA but it is definitely a more robust, post-election tabulation audit.

1 Like

@votinggeek is right. The NM audits are a big step forward, but they don’t guarantee a risk limit, so aren’t RLAs.

Good points:

  • State-wide coordinated auditing of several contests
  • Audits conducted by independent auditor (a CPA)
  • Margin-sensitive audit - puts more focus on close contests
  • Batch-size-sensitive audit: more focus on large precincts (but by registration numbers not actual ballot counts?)

Issues, last I checked:

  • Auditing started before all ballots were counted, thus inaccurate margins
  • Last I checked, under votes weren’t included in the margins they used.
  • Not all ballots were audited, e.g. not provisionals, perhaps not late-arriving absentee, etc.
  • Not clear if the precinct counts are published before the random selection happens, or how to download them
  • They assume that the precinct results aren’t very far off predictions (assuming a maximum 15% “within-precinct-miscount” I think), but I don’t think they have a robust way to monitor or even predict many tallies by precinct.
  • No auditing of contests with greater than 15% margin, so an adversary could steal more votes and prevent an audit
  • No permanent archive of audits that I’ve found, and old URLs stop working, e.g. http://www.sos.state.nm.us/uploads/files/2016PostElectionVotingSystemCheckAudit.PDF
    For reference, I put an archived copy of that one at
    http://bcn.boulder.co.us/~neal/elections/NM-2016PostElectionVotingSystemCheckAudit.PDF

So again, NM audits are much better than the vast majority of states, but had significant gaps last I saw.

1 Like

I would call the New Mexico Audit a risk based audit. It does not meet the definition of a risk-limiting audit, but it does require a larger sample for smaller margin contests. For more information about New Mexico Audits you might want to contact Paul Stokes.