Instructions for observing post-election audits

Purpose of the post-election voting-machine audit

Federal law requires states to audit their voting machines after at least some elections to see whether they count votes accurately. The state law governing our post-election audits, s.7.08(6), Wis. Stats, fulfills only the bare minimum for the federal mandate and has several other flaws that prevent these audits from conclusively verifying election results. Nevertheless, if the audits are done well, they can provide local officials and citizens with useful information and might detect any serious miscounts.


UPDATE (and GOOD NEWS!): At its October 28, 2014 meeting, and at the urging of WGN's Election Integrity Action Team, Wisconsin's Government Accountability Board corrected one of the biggest flaws in our post-election audit process. They voted to amend their previous instructions that clerks wait until after all deadlines for recounts had passed before they could begin the audits, and replaced it with instructions that local election officials could begin post-election audits AT ANY TIME.

Thanks to GAB Board member Hon. Elsa Lamelas, who made the motion to amend the instructions; Hon. Harold Froehlich, who seconded it, and Hon. Thomas Barland and Hon. John Franke, who supported it. (Hon. Gerald Nichol was absent from the meeting.) Only one judge voted against the change, and all their questions and remarks evidenced serious and sincere concern for accurate election results.

These observers' instructions have been revised to reflect the clerks' new freedom to verify their voting-machine output whenever they see fit.

The audits also have an important quality-assurance purpose. Hectic activity on Election Day leaves little time to calmly assess what is being done well or sloppily. Post-election audits provide opportunity for election officials to review the quality of some election practices.

Why observe the audits?

Without observers, transparency cannot be achieved. Election officials cannot demonstrate anything to the public if the public isn’t present.

Even if observers say nothing, their presence helps to ensure officials perform the tasks thoroughly and in accordance with instructions, and to ensure that any problems discovered in the audit are not dismissed or ignored.

Observing hand counts (at least those not part of a contentious, hurried recount) can be a moving experience. People have fought and died for the right to count votes; to be present as our neighbors carry on this sacred heritage is an honor. 

Finally, because voting-machine audits are often relatively relaxed, unhurried occasions, they provide opportunity for the municipal clerk and interested citizens to develop mutual respect for their shared interest in accurate election results.

When are voting machine results audited?

If your municipal clerk does no more than the law minimally requires, your election results will rarely be checked for accuracy, if ever.  You and your fellow citizens can request that your municipal clerk conduct audits on his or her own initiative, but if he or she declines, you will need to wait until GAB orders an audit.

After the November elections in even-numbered years, the state Governmental Accountability Board (GAB) randomly selects a very small number of precincts in which the voting machines will be audited—likely only 100 of the more than 3,500 Wisconsin precincts.

GAB orders those municipalities to perform a hand count of the votes in each selected precinct and to compare the hand-count results against the machine-tabulated results in four races identified by GAB to make sure the machines counted correctly. In some counties, the county clerk’s office performs the audits for the municipalities.

The timing of these audits limits their value. GAB instructs municipalities to wait to verify the results’ accuracy until after GAB has certified all statewide election results as final, which is likely in mid- to late November. (You read that correctly: Many states don’t allow unaudited results to be certified as final. Ours recommends it!) The audits can occur anytime after the voting machine output is printed out, before or after the ballot bags have been sealed. Your clerk may not be aware of this change in GAB policy; it was made on October 28, 2014. 

What happens at an audit?

Post-election audits are usually calm; the action will be deliberate and careful. Because these audits are so rare, small municipalities may never have conducted one. Even in larger municipalities, it’s likely only one or two people will have audit experience. Expect the officials in charge and the auditors to pay close attention to the written instructions, even reading them aloud at times.

There are no requirements regarding who the clerk might call upon to count the votes; they usually select employees from their office or experienced poll workers.

On the morning of the audit, sealed bags containing ballots and election records will be brought to the audit site and opened. Conscientious clerks will inspect the material for any signs of tampering and will review the chain of custody.

The auditors will be provided with tally sheets, pencils or pens, and perhaps calculators and white-out. They begin by counting the ballots into stacks of 20, and the stacks of 20 ballots are arranged in batches of 100.

The auditors will pair up, and each pair will take two batches of 100 ballots. Each auditor will then count the votes from one batch and trade batches with his/her partner. When both have counted both batches, they will compare their tally sheets. If their totals match, the two batches will be set aside, and they will move on to the next two batches.

If the two auditors’ totals for any batch do not agree, they will jointly review the ballots and try to find the ones that they counted differently. Experienced auditors will flag those ballots in some way, so that they can be located again in the final reconciliation process. When the two auditors agree on the vote totals in each batch, one will change his or her tally sheet, and they will move on to the next.

When all the ballots in the precinct have been hand-counted, the subtotals from all the tally sheets will be added together and compared against the totals on the voting-machine tape that was printed out on Election Night.

If the hand-counted totals in each race agree with the machine-tabulated totals, the audit is done, and the official in charge completes a report for GAB.

If any totals disagree, the auditors need to try to figure out why. They will need to check their addition, and then most likely, will engage in something of a guessing game trying to figure out how the voting machine might have read ambiguously marked ballots when auditing an optical-scan machine, or questioning their own counts.

When the auditors have decided how to explain any discrepancies, the audit is complete and the official in charge completes a report for the GAB.

How to find out whether and when an audit will take place

A day or two after a November election, call your municipal clerk or the GAB (608-266-8005) to find out whether any precincts in your county or municipality were selected for a post-election audit.  (GAB orders no audits after any but November elections.)

If you decide to observe, contact the municipal clerk to ask about his or her plans for scheduling, but be prepared to cut the clerk some slack: GAB doesn’t make it easy to schedule these audits.

GAB tells the clerks to conduct the audits only after GAB certifies the statewide results, but can say only that certification will happen sometime in mid- to late November. On that day, GAB will notify the clerks the audits can go forward, but the clerks must give the public 48 hours’ notice and complete the audits within two weeks.

Most clerks will try to follow GAB scheduling instructions, but because statutes do not require clerks to wait, a few will schedule the audit promptly. If the clerk intends to follow GAB recommendation, ask the clerk about the best way for you to get timely notification when the audit is finally scheduled.

You are allowed to show up unannounced, but if you let the clerk know ahead of time that you will be observing the post-election audit, he or she will be able to make sure there is room for you, and will be able to let the auditors know to expect observers so they won’t be surprised and nervous to see you.

How to observe a voting-machine audit

1.  Read these instructions before arriving, and follow the general instructions for observers, which are described in a separate document, also available from the WGN Election Integrity website.

2.  Bring a copy of the GAB audit instructions with you. The instructions at that link are from 2012; current instructions will probably be posted on the GAB website shortly after Election Day. New instructions may be available at that time by following links from this page. If you cannot locate current instructions, call GAB at 608-266-8005 to ask where they might be located.

3.  Arrive a little early to allow time for cordial introductions. Explain to the official in charge that the point of observing is to verify that the votes are being counted correctly, so you would like to be able at least part of the time to look over the auditors’ shoulders in a way that will not disrupt their work. At a minimum, you will need to see any ballots that the auditors discuss. The clerk should be able to accommodate these requests, as they do during recounts.

Auditing a single precinct can take all day, depending on the number of people counting votes, so you may want to plan to be there only at the beginning and at the time they reconcile the completed hand count with the machine tape. They will be unable to predict exactly when they will finish the hand count, but you might be able to estimate that after watching them count the first few batches.

4.  Although the clerks’ instructions do not specify this, the audit should begin with an inspection of the election records, including the ballot bags, to make sure all necessary records are on hand and are in order, much as the clerk would do for a recount.  The marked ballots or audit paper trail should still be sealed in bags or containers in such a way that no ballots can be inserted or removed without breaking the seal, and that show no signs of tampering; the inspector’s statement should be on hand and in order, containing the seal numbers that match those on the ballot bags, among other information; and the machine-printed results tape should be on hand.

If it appears that the machine tape or the sealed ballot bags have been tampered with before they are opened for the audit (unless there was a recount), please follow the instructions for when you see a problem, below.

5.  As the audit tasks proceed, ask questions throughout, as long as your questions don’t interfere with the purpose of the audit or slow it down more than the officials are willing to accommodate.

6.  Issues you should know about (Touch-screen audits)

Unusable voter-verifiable paper audit trail - Wisconsin law wisely requires touch-screen machines to create a voter-verifiable paper audit trail (VVPAT) –a paper print-out of each ballot cast by each voter. Although each voter has an opportunity to view this print-out and to cancel and re-cast it if necessary, research has shown that few voters ever look before they leave the voting booth. If the tape is malfunctioning---failing to advance, failing to print legibly, etc.— many voters can cast their ballots before any one notices and reports the problem to election officials.

If the audit finds that the paper trail cannot be read, the audit is impossible, because there is no record to audit. Follow the instructions for what to do if you see a problem, below.

Cancelled votes - Experts on election fraud have determined that, if a malicious programmer does manage to manipulate a touch-screen voting machine, the fraud could leave evidence in the form of an unreadable paper trail (see above) or a larger-than-expected number of cancelled ballots.

GAB’s instructions for post-election audits do not require clerks to be alert for such evidence, but as an observer, you can be. Because no one in Wisconsin looks for this, it’s hard to say how many cancelled-and-recast votes can be expected due to predictable human error or voters changing their minds at the last minute.  We recommend that if you see more than two cancelled ballots out of every thousand, you ask the auditors to examine two things more closely.

First, ask them to note to confirm that the number of valid ballots cast (not counting the cancelled ballots), equals the number of voters recorded in the poll books as having voted.  If the number of valid ballots cast is less than the number of voters, the cancelled ballots indicate a voter entered the voting booth, cancelled the vote entirely, and left without voting at all—something that should be extremely rare.  If the audit discovers that this occurred in more than two votes out of every thousand, please follow the instructions for reporting a problem, below.

Second, if the number of votes cast equals the number of voters recorded in the poll book, ask them to note whether all the re-cast votes were for the same candidate. (They can tell which votes were changed by comparing the cancelled ballot to the ballot immediately following it on the tape.) A pattern in which all the cancelled-and-changed votes were for the same candidate is consistent with a non-random malfunction.

Follow the instructions for reporting a problem, below, if the total number of cancelled-and-switched changed votes is more than two out of every 1,000 (0.02%), and the changes favored the same candidate. Suggest that the municipal clerk review the inspector’s report for that precinct to see whether it noted any voter reports of errors on the paper trail on Election Day.

7.  Issues you should know about  (Voter-marked paper ballots)

Ambiguously marked ballots   A certain proportion of votes can be expected to be ambiguously or sloppily marked, so that the hand-counters will disagree on the voter’s intent. 

This is not as serious an issue in audits as it is in recounts, where a few votes could change the outcome. Nevertheless, watch to see about how often the auditors disagree, and how they resolve their differences. Ask to see the ballots that seem to them to be ambiguously marked, so that you can develop your own sense of how well the voters marked their ballots. A very large proportion of ambiguously marked ballots—more than an average of one in every 100 ballots—may indicate a problem you could discuss with your municipal or county clerk: Could the ballots be better designed next election, could the voters be instructed more effectively?

Machine-unreadable votes   A subset of oddly marked ballots are marked in such a way that humans can easily discern the voter’s intent but a machine cannot.  For example, a voter may circle a candidate’s name on the ballot rather than filling in the target dot beside the candidate’s name, make a very faint mark, or use the wrong kind of ink.

A large number of these ballots—more than one in every 200 votes, according to some experts—could indicate a serious problem, which the auditors may or may not notice depending upon how closely they follow the GAB’s instructions.

Curiously, GAB has in previous years instructed the auditors to “count the votes as the machine would have counted them,” an instruction that could defeat the purpose of the audit if followed too literally. For example, if one voting station had been equipped with a colored pen that the optical scanner cannot read, every ballot completed with that pen would not have been counted on Election Day, although the voters’ intent would be clear to human eyes. If the auditors follow GAB’s instruction to “count the votes as the machine would have counted them,” they too would ignore the votes and mistakenly conclude the machine-tabulated results were correct.

Fortunately, common sense leads most auditors to count those votes, so that the issue becomes apparent only when the hand-counted results are compared to the machine-counted results.

If the auditors you observe seem to be altering the hand-count results in favor of their guess at how the machine counted in more than one vote out of every 200, follow the instructions below for what to do if you see a problem.

8.  When the hand-count is complete, the results are compared to the machine-tabulated results. If the two totals match, the machine tabulations are confirmed accurate and the audit is done. The clerk will complete a report to GAB indicating an ‘error rate’ of zero.

If the two totals do not agree, the clerk will need to figure out why they differ before he or she can calculate an error rate, per GAB instructions. Common human-error mistakes include addition errors and forgetting to exclude late-arriving absentee ballots (which don’t show up on the Election-night machine tape).  

This is another point at which the human hand-counters might notice that they were able to discern voter intent better than the optical scanner for some ballots, and can explain small discrepancies between the two counts by reference to several oddly marked ballots. This is no cause for alarm as long the the discrepancy between the two counts is less than one vote in every 200 or so.

If the hand-count totals need to be adjusted by more than that, there may be a problem. There may also be a problem if the auditors settle on an explanation for the discrepancy that involves a systemic issue with the machine or how it was set up. For example, if they conclude that the machine was unable to read all ballots created with an AutoMark device used by voters with disabilities, this issue should be reported regardless of how many votes it affected.

When the auditors have settled on an explanation for the discrepancy, the audit is complete. GAB instructions require them to disregard any discrepancy they can explain, and to record a calculated error rate of zero.

What to do if you see problems

If you see any practices that don’t match the instructions, ask the clerk to explain. There may be some necessary, harmless variations the clerk is doing for a sensible reason. As long as the main purpose of the test is fulfilled—verifying the accuracy of the machine-tabulated total by comparing it to the results of an objective hand count—there’s no need for observers to be sticklers.

If the clerk cannot give you a good explanation for the variation, and you think the variation prevents the audit from verifying the results’ accuracy, refer the clerk to the requirement or instruction that you believe is not being followed. Explain the reason for your concern.

If you cannot work out a mutually satisfactory resolution with the clerk, encourage him or her to contact the county clerk about the matter, and say you will do the same. Promptly write down the relevant details of the problem as you observed it, clearly, factually, and completely. Then call or email the county clerk to report the problem and ask to be kept informed as the problem is resolved.

If you believe the problem may have prevented the audit from discovering and documenting any serious miscount that might have been present, contact the Elections Division of the Wisconsin Government Accountability Board. GAB can be reached by phone at (608) 266-8005, or email at [email protected]. Consider completing a complaint form ( ), and faxing it to GAB at (608) 267-0500.

Please also consider sending an account of the problem to the Wisconsin Grassroots Network at [email protected]. We will be compiling data on the problems noted by observers to improve understanding of election integrity in Wisconsin.

- - - - - - - - - - -

A printable pdf version is this document is available through a link on this page.

These instructions can be improved by your feedback. Please comment below or email your suggestions to us at [email protected] We particularly invite suggestions from observers who have used these instruction at a poll-closing, and from election officials including people with experience as elections inspectors.

We would very much appreciate hearing about what you observed, good and bad. Please email us at [email protected] or comment below to tell us how it went!