Correction to earlier blog post: Are our election results ever audited?

(You are reading a blog post on a legacy website that is no longer actively maintained. Visit WisconsinElectionIntegrity.org for current information and discussion.)

A few days after I posted a letter to the chair of the UW Statistics Department here, in which I asked for a volunteer to help develop efficient procedures for verifying Wisconsin’s voting- machine output, I received a cordial email from David Buerger, Elections Specialist at Wisconsin’s Government Accountability Board. He took issue with several statements in that letter, and I appreciate his willingness to communicate and to speak up. I replied that I would review his comments and update this blog if I saw anything that merited correction.

The entire text of his remarks is here but for brevity’s sake, I’ll summarize. Buerger objected most strongly to my statement that “Our election officials never check to see whether our voting machines produce accurate counts,” except in the case of a recount. Buerger characterized that statement as ‘patently false.’

Strong words, but he is correct that once every 2 years in 4 races in about 100 precincts (out of about 3,500 in the state), GAB orders local election officials to verify Election-Day voting-machine output, under s.7.08(6), Wis. Stats. Results of the audits that followed the November 2014 election will be released in March 2015, more than three months after the unverified output was certified to be Wisconsin's final official election results, and more than two months after the winners identified by that unverified output were sworn into office.

Therefore, a more precise and technically accurate sentence would have been: “Our election officials never check to see whether our voting machines produce accurate counts in a way that ensures no electronic miscounts affect our election results” except in the case of a recount.

To illustrate the difference in meaning between those two sentences, imagine that Buerger had said, “McKim never proofreads her letters.”

If I responded by saying, “That’s patently false. I proofread about 3 percent of my letters--if you order me to--about a month after I mail them,” would you have confidence in even one of my letters? Of course not.  Looking for errors is a meaningful exercise only while errors can be corrected.

The audits that GAB orders under s.7.08(6), Wis. Stats., are better than nothing. I’m glad they are done. I don’t want to keep pointing out their shortcomings. Really I don’t.

But no one should perceive those audits as a safeguard against electronic miscounts being certified as final results in any election, ever.

We pointed out the limitations of those audits in a report in July 2013; since then the Legislative Audit Bureau described some of the same issues. So, at the risk of boring all but the nerdiest of readers, I will again and I hope for the final time list some of the reasons why these audits do not safeguard a single Wisconsin election outcome, ever.

  • Too late: GAB allows the audits to be performed after the election results have been declared final, when there is no way to correct any miscounts that might be discovered.  
  • Too few races in too few elections: GAB orders audits of voting machine output in only four races on the November ballot. Referenda, purely local races, and all contests in Spring elections, Fall primaries, and special elections are never audited.
  • Sample size is not designed to verify outcomes: GAB orders audits in only about 100 of the state’s 3,500 or so precincts—a sample size not intended to be sufficient to confirm the correct outcome in any race, either local or statewide;
  • Sample selection is not designed to verify outcomes: The statutory purpose of the audits is to check to see whether the specific models of hardware used in this state are capable of producing correct results, and the sample is drawn with an eye to the type of hardware used, not outcomes verification.
  • Definition of ‘error’ defeats the purpose of verification – GAB instructs auditors to ignore voter intent as they count the marks on the ballots, but instead to count the votes as the machine would have counted them; and
  • Lack of procedures to expand the audit when errors are detected – Neither statutes nor GAB instructions provide for expanding the audit beyond the single precinct where an error was found. So even if a timely audit were to detect a mis-programmed machine in one precinct, there is no provision for auditing the results in other precincts even where machines were programmed by the same company or office.

And for what it’s worth, the pre-election voting machine tests that Buerger also cited, when performed correctly, are useful for detecting only one type of voting-machine error (incorrect set-up), but are of no use in protecting against or detecting miscounts due to machine malfunction or deliberate electronic election theft.

I won’t repeat the rest of Buerger’s points, on which he and I are in substantial agreement: Budgets are tight; election laws are in turmoil; everyone works hard; no one is stupid; GAB staff decided against developing effective, efficient audit methods due to workload considerations—all true.

And heck, I’ll go even farther than Buerger in defending the GAB: The Board took a big step forward in October 2014, when they voted to make it clear that local officials are permitted to perform timely verification of voting machine output, a practice their previous policy had explicitly discouraged.

In their further defense, it’s useful to remember that accurate election results are not the GAB’s responsibility anyway. Statutory responsibility for the election results' accuracy rests with the county boards of canvass, and to a lesser extent with municipal election officials and the citizens who could hold them accountable.

Responsible citizens and local elected officials have all we really need from GAB. We know that unverified computer output is routinely used as final election results in every Wisconsin election except those that are recounted, and we know that practice is unwise, imprudent, and contrary to the recommendations of every national elections-administration expert.

It’s long past time for citizens and local elections officials to start working together to improve local management of our elections technology. That’s not GAB’s job. It’s ours.



Showing 1 reaction

Please check your e-mail for a link to activate your account.
  • commented 2016-12-22 14:10:48 -0600
    An additional explanation, relating to the instruction that the municipal auditors are to ignore voter intent and “count the votes as the machine would have counted them.”
    The intent of this instruction is consistent with the purpose that state election agency staff see: the audits are to confirm the voting machines themselves (not their programmers, not their operators) is still capable of counting correctly. In their minds, they consider misprogramming to be human error (or fraud), not machine error, and therefore NOT what these audits are looking for.
    I checked this interpretation of their intent with them, asking how they believe the municipal clerk should have handled the audit had the City of Medford, in November 2004, been selected for an audit. In that election, that city’s machines had accidentally been programmed to ignore all straight-party ticket votes, effectively disenfranchising one-third of Medford’s voters.
    They confirmed that the audit should have produced, under their instructions, a finding of “No Error,” explaining “You can’t blame that on the machine.”