The Racine County Citizens' Audit is ON! Sign up to help count votes, using this website.
Volunteers will be counting votes at the Racine County Courthouse, 720 Wisconsin Avenue, Racine on Tuesday November 14 and Wednesday, November 15. An additional shift will be added on Thursday morning if necessary.
Vote-counters will work in three-hour shifts, starting at 9 AM and 1 PM each day. Volunteers can work for more than one shift if they want.
The task will be easy, fun, and rewarding. A county employee will display each ballot using a document projector, so that many people can see the same ballot at the same time. You’ll sit at the county board conference table and use a hand-held mechanical clicker to count votes for only one candidate. You'll be given frequent opportunities to compare your totals with another counter doing the same thing.
We're paying for this, so volunteers will be treated with respect--unlike at last year's recount.
If you cannot stay for three hours to help count the votes, drop in anytime to observe. You will see how your county clerk could perform an efficient, transparent hand count at any time he or she wants--such as before declaring election results final.
Having observers come and go all day will also demonstrate citizen demand for better election verification practices. So please stop in! Or sign up to help count votes, using this website.
This citizens' effort is necessary, because Wisconsin elections need, but do not have, routine manual verification of all election results. In both the election and the recount last year, the Racine County Board of Canvassers certified computer-tabulated vote totals that were obviously incorrect. In one precinct, the machine said that 1 in every 12 ballots contained no vote for president! In doing that, they disenfranchised every voters whose ballot had been wrongly read by the machines--we estimate more than a thousand.
Blind trust in voting machines--new or old--cannot continue. In this day and age, it is crazy to trust any computer's output without checking--and vote-counting computers are the most attractive targets of any.
Please visit this website to sign up! And tell a friend—observers will be welcomed.
September 27, 2017 --
Main point: Public officials must keep the public both safe and calm.
The danger is when election officials' goal is reassurance, not safety.
Yesterday’s Wisconsin Elections Commission (WEC) meeting was packed with more cameras than I’d ever seen there. A few days earlier, the federal Department of Homeland Security announced that Russian-government backed hackers had tested the security of Wisconsin’s online voter registration system. They hadn’t gotten in. The ‘attack’ was, the computer experts say, like jiggling a locked door knob.
What voters get: “As you can see, it's a beautiful day, the beaches are open and people are having a wonderful time." - Mayor Vaughn, to a reporter.
“I don’t get it.” I told a reporter as the meeting got under way. “What's the news here? Hackers are continuously testing every computer system. The Russian government is known for cybercrime. It would be news if they were not testing the security of our elections systems.”
I don’t remember his response, other than it wasn’t convincing. I fear the real answer is that his editors know which stories get the web clicks.
The facts that WEC shared were as I expected. State officials from the WEC and the Wisconsin Division of Enterprise Technology (DET) explained their system of continuous defense against hacking of our voter registration system (which is separate from the tabulation system, also known as the voting machines). Millions of efforts to get into the registration system are detected every week, from anonymous Internet addresses all over the world. Unrecognized addresses are locked out and if that fails, any unauthorized changes will be promptly noticed and reversed. If that fails, daily backups are made so that if some malicious code ever causes the system suddenly to garble or erase our voter registrations on election morning, a correct version can be quickly brought up. If that fails, paper backups are printed immediately before each Election Day.
State officials were convincingly competent and straightforward. The story that later appeared in the paper made the federal officials, not the state ones, look like the Keystone Cops.
WEC and DET took the opportunity to explain the security of our voter registration system to the press—while the press was willing to listen. When officials are keeping us safe, reassuring the public is usually as easy and effective as just telling the truth.
The officials’ explanation about our voter registration system confirmed my trusting assumptions about its security.
But the security of our vote-counting software is a completely different story.
Our election officials' silence about security for that system should be a dead giveaway there's a shark in the water.
Like ‘baby’ in a pop song, election officials’ yesterday continuously repeated “We’re talking about the voter-registration system, not the vote-counting systems.” The reporters’ keyboards clicked along to the beat. Yeah, yeah, yeah. None seemed to notice the story within that silence on the vote-counting software.
Here's why we don't get convincing, impressive descriptions of the security system for our voting machines: Because it doesn't exist. At least when sharks are eating tourists, someone notices. But if anyone is hacking our voting machines, their crimes would go undetected as we swear their chosen victors into office.
Reassuring spin: "We've seen no evidence of tampering with the vote-counting system." The furor about Russian testing of our voter-registration system’s security was made possible by federal officials’ looking for it. No one--local, state, or federal--reviews Wisconsin's election results to make sure they are accurate. None of them make any efforts to detect any doorknob jiggling of our vote-counting software, which is proprietary and controlled by the voting-machine companies.
Reassuring spin: "Our decentralized vote-counting system makes hacking unlikely." After the vote-counting software is produced at the companies, it's downloaded to the dozens of computers that will be used to design the ballots for each election and to tell the voting machines how to read those ballots. These are the 'election management systems' that reside at the vendor's regional offices, the voting-machine service companies like Command Central, and in the offices of county election officials.
When election officials talk about the security of the vote-counting systems, they often refer to this decentralization. They say it makes the system harder to hack.
But they cannot possible imagine that, to tip a statewide race, a hacker would need to design a hack specifically for every type of voting machine used in Wisconsin and alter the results in every county. You can see the silliness of that--What's Russian for "Darn it, we missed Forest County. Well, maybe next year."? There are enough votes in Milwaukee County alone, or a few other counties, to control the outcome of most statewide races.
Not only does the decentralization provide little protection, it multiplies the possible entry points and places them in the physical control of an army of people with no particular IT security expertise, and often no access to any.
After the software is downloaded to the local election-management computers, it's revised for each new election and then copied onto removable drives--typically, the same sort of USB drive you can buy at the drugstore. The drives are then handed off to the municipal clerks, who load the software onto each voting machine.
On Election Day, it's in the physical control of the poll workers. At this point, we should probably be hoping that the possessors of the software have no IT expertise, rather than wishing that they did.
Between elections, the vote-counting computers are stored in very town, village, and city in the state, under conditions that the election officials themselves don't always control.
No one exercises any oversight of this disjointed system. Computer security expert Bruce Schneier told NPR's Science Friday that federal voting-system security standards were outdated long ago, and no one is now exercising any oversight even if the standards were current. Vendors can coach county clerks on how to maintain security, but they have no way of knowing whether the clerks follow their instructions. To my knowledge (and I asked when I can), no state or local official ever attempts to oversee or even ask about voting-machine company security. They wouldn't know how to evaluate it if they did, or any authority to force corrections.
Johns Hopkins University Computer Security Professor Aviel Rubin made a point of contacting the major voting-machine companies who count America's votes. He reported "I have yet to meet an American voting system manufacturer that employs even one full-time trained expert in computer security."
Reassuring spin: "Our voting machines are never connected to the Internet." This used to be true, but there's no machine on the market anymore without the capability of electronically transmitting results after the polls close. That, however, is not and never was the big risk. Connecting a voting machine to the Internet or to a cell phone tower after the polls close doesn't give a hacker any opportunity to alter a hard-copy poll tape you've already printed. Having observed more poll-closings than I can count and several canvass meetings, I can vouch for the fact that is the one hack our election officials would likely detect and could easily correct.
The vulnerability comes before the votes are counted, not after. The big risk of manipulation--in fact the one that forensic IT security experts deem the greatest--doesn't come from the Internet at all, but from insiders with authorized access to the software. Because no state or local election officials have the authority or ability to inspect the vote-tabulating software for integrity, even lightly sophisticated individuals--at the voting machine company, the service company, the local official's office, or anywhere along the chain of custody--could alter the software and not be noticed. Thousands of people have authorized access to our vote-counting software or hardware between every election. Many of them, in the testing laboratories, voting-machine companies and service companies, understand the code. Many of the others likely can be bought--they are humans.
But hackers without authorized access can get in. The vote-counting software is created, updated, and maintained not on each individual voting machine, but on computers that are almost certainly, at some time, connected to the Internet.
And local election officials have no way to tell whether and when the individual voting machines are communicating with other machines. Wireless communications capability can be installed inside any computer or voting machine--antenna and all--without their knowledge and controlled by anyone within transmission range. Local election officials never inspect the insides of the voting machines for surreptitiously installed wireless cards, and few would know what to look for if they did.
Reassuring spin: "No election has ever been hacked." The truth is, our election officials wouldn't know if one had. They don't use the one practical opportunity--checking the results against the paper ballots--to check the system's integrity. If any election ever has been hacked, it's likely no one noticed.
What voters need: “Smile, you son of a bitch.” – Martin Brody, to the shark.
Yet despite the widespread concern about the security of last year’s presidential election, not a single state had routine procedures in place to verify an accurate statewide vote count. Michigan, Pennsylvania, and Florida proved unable to document accuracy even when directly challenged, unable to get a recount even started.
Wisconsin did best. Every county at least double-checked things like the handling of absentee ballots, but only half of the vote totals were checked for accuracy. The other half were just run back through the same computers, so any electronic miscounts would have just been repeated. We know that some were miscounted twice.
State officials in Wisconsin recently scored a first, when in January they detected a few miscounting computers—after the winners from the previous November were already sworn into office. To their credit, they decertified the machines. They are still are not sure what caused the miscounts—they know ink color on the ballots contributed, and that from their size and randomness, the miscounts seem unlikely to be even a trial-run hack.
What to do?
Face it: State and local election officials will never have the authority, skill, or money to maintain strong IT security for our vote-counting software. It's just not going to happen. Elections are too intermittent, the workforce too temporary, the property taxpayers too stingy to make good security possible.
Our only hope for protecting our election results from hackers--and from malfunctions, glitches, and human operator error--is to notice and correct any miscounts before results are certified.
If the polls opened and voter registrations were garbled, we would notice. Perhaps that's why those responsible for the software are so vigilant--they know any laxity will get found out.
But we cannot sit by the television on Election Night and say “Hey! That’s not how we voted!” Voters have no way to tell honest election results from false ones. And maybe that's why checking accuracy is such a low priority for our election officials. If they don't detect the miscounts, they can keep saying--honestly--"We've never known an election to be hacked."
Most states now have paper ballots, or at least paper audit trails, that could be used to check the accuracy of the computer output. National election authorities have developed, and national officials endorsed, efficient methods that don't require a full hand count.
Every other public official takes responsibility for the accuracy of their work product. It's long past time for voters to insist their election officials do the same.
September 26, 2017 -- There’s good news and there’s … no-worse-than-usual news.
The good news is that today, the Wisconsin Elections Commission did what no Wisconsin elections agency has done since the introduction of computerized vote tabulation: They decertified a voting machine, the Optech Eagle.
And they did it for the best of reasons: It wasn’t counting our votes reliably. Now that so many ballots are marked in voters' homes, in all sorts of ink, the machine is "no longer meeting voters' and officials' expectations."
This is good. Not perfect, but good. The vote was unanimous. The Commissioners didn't debate whether the machine should be decertified, but how quickly. They didn’t vote to decertify immediately, but they soberly considered that possibility. And they did adopt some immediate safeguards.
As of today, all municipalities using the Optech Eagle must either count mailed-in ballots by hand, or re-make (that is, copy over) them using ink that can be detected by the machines. And they must keep doing that until they replace the machines, no later than December 31, 2018. In addition, if any contest tabulated by an Optech Eagle is recounted, it must be by a hand count.
This decision had several good angles to it.
First, the Commissioners’ comments, specifically mentioning Racine County, indicated that they accepted as true the reports of Liz Whitlock and the other observers during that recount, even though Racine County officials have not yet acknowledged any problems.
Granted, it would have taken chutzpah for the Commissioners actively to deny that Racine miscounted both the election and the recount, given all the hard evidence of similar miscounts from other counties and the weirdly high undervote rates that county's canvass signed off on.
But the culture of election officials, in Wisconsin and elsewhere, is to band together against concerned voters and, if not to deny their truth, at least ignore it. But in the discussion today, I did not pick up one whiff of the get-these-troublesome-citizens-out-of-here attitude to which we’re so accustomed. Good work, Liz and the rest of the Racine team! The State hears you, even if your clerk doesn't (yet).
Second, the instruction that any recount be conducted by hand shows more courage and commitment than I’ve seen from any public official in a while. Here’s why: If some county decides to contest that requirement, it’s likely that a court would decide that the WEC has no statutory authority to order a hand-counted recount. The Commissioners were aware of that when they voted, but went ahead anyway. Set aside the fact that recounts are a thing of the past in Wisconsin; I like the kind of leadership that says, “Let’s do the right thing and see if anyone tries to stop us.”
Third, the decision signifies a distinct break from the old Government Accountability Board’s attitude toward elections technology. The old GAB—both board and staff—seemed resistant to even the idea that voting machines could miscount. I remember talking with them about the Medford miscount, when misprogrammed machines ignored all straight-party ticket votes. About a third of that city’s votes in a presidential election were lost. GAB staff told me, with a pained expression, “You can’t blame that on the machines!”, as if I would hurt the machine’s feelings if it heard me say it needed to be audited. I can so easily imagine the old GAB Director Kevin Kennedy defending the Optech Eagle with such an argument.
But WEC Director Michael Haas and the Commissioners are willing to take a stand: A machine that cannot count a valid vote has got to go.
I know I may be giving WEC credit for understanding the obvious, but it is a change from the way they were talking only two years ago. And that's good.
The no-worse-than-usual news is, well, unsurprising.
- The staff analysis of decertification stressed cost and convenience for clerks above all other considerations—to the point where I sat there seriously trying to think of how we could frame the risk of election fraud as a cost issue.
- No one ever has investigated or resolved the causes of the worst Optech Eagle miscount. The WEC is just guessing it was the wrong ink. In Marinette, three voting machines missed 9.6%, 26.5%, and 30.8% of the votes on the ballots they processed. It's almost certain that ink had something to do with it, but if the voters marked their ballots at home, why did voters in one section of town use unreadable ink at more than three times the rate of another part of town? And to add to the mystery, the municipal clerk told me that most of the absentee ballots in all three precincts were in-person early voters who marked their ballots in her office. Why would she provide the wrong pen at all, never mind provide it at different rates to the voters from the different precincts? Finally, in the one municipality where WEC staff did do a serious investigation of the cause of an Optech Eagle miscount, they couldn't pin it entirely on ink. Something else is going on with those machines, and remaking the ballots might not fix it.
- Director Michael Haas, on at least two occasions today, referred respectfully to our testimony, and clearly understood what we were saying. But when he spoke most directly to the prospect of future routine election audits, he called it a ‘legislative issue.’ To me, that revealed his perception that Wisconsin’s local election clerks will not agree to verify election results unless forced by law. He’s probably correct, but that’s pretty darn sad. Thank goodness few other public officials take the same attitude toward their work product.
- In my oral testimony, I cited several instances in which county boards of canvass certified obviously incorrect vote totals. I also spoke of the hard fact that none ever verify the vote totals before they certify. Sure enough, like a patellar reflex, the municipal clerk who spoke next offered indignant testimony: “We do too care about accuracy,” though she offered no facts to back up that claim.
The truth of her statement depends on what she means by ‘care.’ I don’t doubt that she “feels concern or interest.“
But until she routinely verifies the vote totals before certifying them, she does not “exercise serious attention or effort to avoid damage or risk.”
So, WEC's attitude toward election accuracy is improving. But the local election officials still haven't mastered Step 1: Accept that you have a problem.
Inaccurate preliminary vote totals are not a problem--if miscounts
are detected and fixed before election results are declared final.
The problem is, as these incidents show, that Wisconsin's local election officials
are making no effort to detect and correct even obvious miscounts.
Have the hackers noticed that yet?
* * *
September 21, 2017 -- Following December's presidential recount, when county clerks reported corrected vote totals to the Wisconsin Elections Commission, they changed at least 17,681 votes from the totals they had previously reported.
Ward by ward, candidates sometimes got more votes, sometimes less. We don't know the full number of miscounted votes, because if recounters both subtracted and added votes to one candidate's total in a single ward, only the net change showed up in the revised totals. Election researchers at the UW-Madison, Harvard, and MIT worked with the data and estimated that in the original results, more than 1 in every 170 votes had been miscounted. *
If you can think of a way to miscount a valid vote, it's likely that somewhere a Wisconsin vote was miscounted like that.
The recount discovered hundreds of absentee ballots still in their envelopes, uncounted on Election Day.
Write-in votes had been treated with extraordinary carelessness. The recount discovered that 1 in every 7 of Evan McMullin's voters had been disenfranchised in the original count. Typos had erased nearly half the votes in precincts in Oneida and Milwaukee Counties. Votes had been double-counted in Eau Claire County.
The miscounts in Marinette, Outagamie, and Racine Counties are of particular interest to those who wonder whether Wisconsin's local election officials are attentive, prudent IT managers.
Those counties and others used a voting machine called the "Optech Eagle." This was once the workhorse of Wisconsin elections, but it never could read votes unless they were marked in an ink that contained carbon.
It didn't read non-carbon ink in 2016, either, when tens of thousands of voters submitted absentee ballots they had marked someplace other than a polling booth equipped with an approved marking device.
So on Election Night, results indicated that Racine County voters were weirdly uninterested in the presidential race. Across the rest of the state, only about 1 in every 130 voters (0.77%) left their ballots blank for president. But in Racine County, almost 1.8% of the ballots were counted as if they were blank. In the City of Racine, the rate was even higher--2.6%. In individual Racine County precincts, up to 1 in every 12 ballots was counted as blank.
In Outagamie County, too, Optech Eagles were telling election officials that dozens of voters had cast no vote for president. In one municipality there, the machine saw no vote for president on more than 1.4% of the ballots.
The machines in the City of Marinette took the prize. The machine that counted ballots cast by absentee voters in the City of Marinette's 7th and 8th Wards printed out results indicating no presidential vote on 9.6% of the ballots. The machine counting Wards 1, 3, and 5 saw no votes on 26.5% of the ballots.
And the machine counting Wards 2, 4, and 6 saw no presidential vote on 30.8% of the ballots it attempted to count.
To their discredit, local election officials either did not notice these obvious errors, or noticed them but chose not to correct them. Officials in these counties signed legal documents attesting that they had reviewed the election results and found them to be "correct and true." But in fact, they hadn't done that at all. At least not until the recount forced them to.
Marinette officials, who conducted their recount by hand, corrected their miscount in the recount. Outagamie officials recounted using the same machines and therefore did not detect their miscount until ordered to do a biennial voting-machine audit in January. By then, it was too late to correct the official vote totals.
Racine County officials still haven't resolved their weird vote totals. They recounted by machine and so certified the bizarre vote totals twice--once after the election and once after the recount--without ever checking to see why they were so weird. In fact, county election officials were so determined to trust unexamined computer output that they refused to check accuracy even after the machines could be seen to be visibly miscounting during the recount.
To their credit, staff of the Wisconsin Elections Commission took steps to figure out what was going on. In January, they asked Outagamie County to send their election materials, including the ballots, to Madison so that staff could examine them and make sure the miscounts were what everyone suspected: an inability to read the ink that many voters used to mark their ballots.
They concluded that ink almost certainly contributed to the miscounts, but that didn't explain all the missed votes. WEC staff concluded “This exercise did not produce a result that allowed staff with confidence to understand how the Optech Eagle treated these ballots.”
Something else, in addition to ink color, was going on, and they couldn't tell what it was.
In June, staff wrote a memo to Commission members, telling them "“The analysis of the performance of the Optech Eagle identified a significant limitation of the equipment.”
The Commission itself then did the right thing, too. They instructed staff to prepare a plan for decertifying the machine, to be decided next Tuesday at their September 26 meeting.
When the Optech Eagle is decertified, it will be illegal to use in Wisconsin. Municipalities will be forced to upgrade to newer machines with enhanced ability to count votes marked in any ink. And that will make our elections a little safer from miscounts. Professor Douglas W. Jones, of the University of Iowa Computer Sciences Department told me: "I've tested the newer machines with everything that I can imagine a voter using. Even glitter pens work, though I still wouldn't recommend them."
We cannot overlook the fact that two failures produced these miscounted election results.
First, the machines miscounted.
Second, local election officials certified those miscounts as "correct and true" anyway.
Replacing the unreliable voting machines with new ones will solve one limited problem. But no sensible person believes that no voting machine will ever again miscount Wisconsin votes. No computer system can do that, particularly when the computers are scattered among every Wisconsin city, village, and town; managed by an army of IT-naive temporary staff; and for which security is a responsibility split among vendors, service technicians, and local election officials.
How many elections did those old machines miscount before the recount revealed the problem to the general public?
And why didn't Wisconsin's election officials notice and correct this problem sooner?
Here's why: Statutes provide Wisconsin's local election officials with a review period following every election, called the 'canvass' during which they can review accuracy before they declare the election results final.
But the election officials don't actually review accuracy. Most clerks are more polite than you see on that Racine County video, but they all place the same blind faith in computer-calculated vote totals. Wisconsin's local election officials just point to the computer print-out and say "Oh, look who won."
On Election Night, when the City of Marinette machines produced such weird numbers, poll workers noted tabulation problems in their Election-Night written reports. Yet neither the city nor county clerks took corrective action. They knowingly certified results that were obviously missing hundreds of votes.
You won't find a county canvass anywhere in Wisconsin that does better. The Dane County clerk will tell you he's the only clerk in the state who checks machines' accuracy after every election (2 machines). But if you press, he will admit he doesn't do that during the canvass. He waits until after he has already legally sworn that the results are accurate and has declared them final.
If the WEC decides to decertify next Tuesday, we commend them for taking action to solve the Optech Eagle part of the problem. We are also asking that they take steps to solve the other half: we are asking the WEC to provide the county canvasses with more specific instructions and stronger encouragement to verify accuracy before certifying election results.
More than 20 other states have already built some kind of verification into their canvass procedures, and they do not declare election results final until they have confirmed accuracy. Wisconsin is no better off than states with paperless touchscreens if all we do is seal our paper ballots in bags and never use them to verify the vote totals.
At the very least, Wisconsin's county clerks need to begin to use simple, routine 'reasonability tests,' which are simple calculations they can do at their desks to look for vote totals that don't make any sense--and then resolve any anomalies they see.
If we have any dedicated county clerks who want to take the lead to bring Wisconsin's canvass procedures into the 21st Century, (ask yours!) our statutes already allow that clerk to adopt whatever canvass procedures he or she wants. National authorities stand ready to help with the implementation of modern methods of election auditing.
There is no reason for Wisconsin voters to continue to trust our right to self-government to anonymous computer programmers or whoever hacked in behind them, or to hope that some IT Fairy Godmother will from now on protect our voting machines from any more glitches.
Our election clerks could be checking accuracy if they chose to--and we need to insist that they do.
* Stephen Ansolabehere, Harvard University; Barry C. Burden and Kenneth R. Mayer, UW-Madison; Charles Stewart III, Massachusetts Institute of Technology - Learning from Recounts, Massachusetts Institute of Technology Political Science Department Research Paper No. 2017-12. July 2017.
"Every vote counts"...even if every vote won't be counted?
Voters and election officials haven’t made up our minds about how accurate we want election results to be.
When it comes to casting votes, we talk as if every vote counts. But when it comes to counting votes, you often hear that every single vote doesn’t really need to be counted.
Miscounts—even large ones—are dismissed as inconsequential if they don’t change the outcome. For example, after the recount, a Wisconsin State Journal headline read: “Recount found thousands of errors, but no major flaws in state election system.”
The reporter, Matthew DeFour, wrote: “At least 9,039 presidential votes weren’t counted correctly on Election Night, and were added to the official results only because of the recount. Another 2,161 votes were originally counted but later tossed out for reasons including to square vote totals with the number of voters who signed the poll book. The more than 11,000 changes to the original vote total represent a minor 0.38 percent error rate out of the nearly 3 million votes counted.”
‘Just identify the right winner’ could be a functional accuracy standard if our election officials ever checked to make sure they did. But Wisconsin election officials don’t, unless someone forces a recount.
In addition, creating an appearance that vote totals are only approximate undermines the “every vote counts” message that motivates people to go to the polls.
Wisconsin statutes contain a more precise accuracy standard
Wisconsin statutes contain a clear Legislative expectation about how accurate election results should be.
Sensible election observers realize a few votes will always be miscounted. A few voters will mark their votes ambiguously; a few ballots will be processed incorrectly. These small errors are, in fact, inconsequential if they don’t change the outcome, but if they do, we want to catch and correct them.
Knowing this, statutes provide for recounts at the taxpayers’ expense if a race was decided by a margin so small that it might have been affected by random errors. But if the margin is so big that the outcome was almost certainly determined by the will of the voters—even if random error remains undetected and uncorrected—the State has no interest in recounting, and the challenger must pay the full cost up front in cash.
When the Legislature revised the recount law in 2015, election officials and others testified that random errors might account for as much as 0.25% of the total votes, but never more. The outcome of any election decided by a larger margin, they testified, would not be altered by a recount because random errors would never be that numerous.
So our statutes now contain an expectation that election officials will produce results that are at least 99.75% accurate, that no more than 0.25% of the total votes are miscounted for any reason—machine or human error.
With that recount standard in the law, it is dangerous to tolerate an error rate greater than 0.25%. If we say to our election officials that it’s okay to miscount (for example) “a minor 0.38 percent” of the votes, we’re setting up a situation in which an election could be decided by random error, yet still not be subject to recount.
Therefore, any miscounts larger than 0.25% of the total vote should be considered serious. Error rates larger than that are worth detecting so that election officials can determine their cause and make sure the error rate is reduced to the current statutory limit.
Whatever the standard, we need to check.
Whether election officials and voters agree on the 99.75% standard written into our recount law, or whether they agree that election officials can miscount all the votes they want as long as they don't pick the wrong winner, we still need to check.
Neither standard means anything if our election officials don't check to see whether they are achieving it.
The thing that deters hackers and crooked insiders is knowing that any miscounts they produce will be noticed and corrected.
The thing that prevents accidental miscounts from doing any harm is catching them and correcting them before election results are declared final.
It doesn't work when election officials lean back and say "Oh, we can check our accuracy if we ever see signs of trouble."
First, they will never see the signs of trouble if they never check.
Second, if election officials wait to audit only when they see signs of trouble, they will run into ferocious resistance from the 'winner' of the suspicious results. "Bias!", they will shout. "You can't question the results in MY election, if you never check the accuracy of any other."
We need routinely to check the accuracy of our election results before they are declared final, and we need to start now.
Pre-election preparation and security
In most of Wisconsin and all of Racine County, votes are read and tabulated by computer. Wisconsin uses only voting machines that have been approved by the federal government, based on reports from private testing laboratories. New and redesigned systems must also be reviewed and approved by the Wisconsin Election Commission. All voting machines used in Wisconsin must use or retain a paper record of each ballot.
Several different commercial systems are used. In November 2016, Racine County used two: The Optech Eagle, an opscan machine that reads paper ballots marked by voters, and the AVC Edge, for voters who need an accessible system. Although an Edge unit is in each of Racine County’s polling places, the large majority of voters (98.2%) chose to use a paper ballot, according to data submitted by Racine County municipalities to the Wisconsin Elections Commission (WEC).
Responsibility for the security for the machines and their software changes several times during each election cycle. When the machines or software are manufactured, serviced, or updated, private companies control the security. Between elections, municipal clerks are responsible for the security of the voting machines themselves, and typically county clerks responsible for the security of the software. Soon before each election, the software is provided to the municipal clerk, and on Election Day, poll workers in each polling place must ensure security.
Before every election, each voting machine must be set up to process the unique ballot for that election, typically by the county clerk or a vendor. During the week before the election, the municipal clerks test each machine publicly to demonstrate that they are counting accurately.
Election Night vote-counting
On Election Night, votes are tabulated by the computer in each voting machine. After the polls close, poll workers, known as election inspectors, unlock the machines and press a “Print Totals” button. The machine then prints a poll tape showing the machine’s calculated totals for each race. Totals are reported for each candidate and for the number of ‘undervotes,’ or ballots from which the machine detected no vote in that race.
Elections inspectors also look through all the ballots to find the write-in votes, and hand-tabulate the votes for registered write-in candidates. These candidates are not listed on the ballot, but did file paperwork with the WEC indicating their candidacy. Election inspectors in Wisconsin are not required to count any other write-in votes, either individually (e.g. 3 for Bernie Sanders, 4 for Aaron Rodgers, etc.) or in the aggregate (e.g., 7 unregistered write-in votes). However, the reporting software and forms allow local election officials to report the total number of protest write-in votes in a category called ‘scattering.’ This reporting is optional, and Racine County did not report these votes. Counties that did count and report ‘scattering’ votes reported that about 1 in every 100 ballots (0.955%) contained a ‘scattering’ vote for president.
The lead election inspector in each polling place completes an ‘inspector’s report,’ on which he or she records basic information about the conduct and results of the election in that polling place, and in particular notes any oddities or irregular occurrences.
Review and certification of election results
Municipal reviews of the election records and results, called canvasses, must be completed within the week following each election. Led by the municipal clerks, these efforts review the inspectors’ reports to note any unresolved issues, resolve any provisional ballots, review the election records to ensure they are complete and in order, and certify (that is, approve as ‘complete and true’) any municipal races on the ballot. (In November 2016, there were none.)
Municipal clerks then submit the election records to the county board of canvassers. The county board of canvassers consists of representatives of the Democratic and Republican parties and the county clerk. County clerks are elected constitutional officers with substantial independent authority.
Statutes describe no specific instructions or procedures for the canvass methods, but do impose timelines. For the most part, the county canvass must be completed within three weeks of Election Day. It is during this period that discrepancies could be noticed, if canvassers looked for them. County canvasses have the authority to have municipalities verify returns that don't make sense. The county canvassers review the returns submitted by the municipalities and if satisfied of their accuracy after review, certify the election results for federal, state, and county offices by signing a statement attesting that the results are 'correct and true.'
Certification makes election results final, and no further review is done unless someone successfully petitions for a recount within a short time after certification. Results are then submitted to the WEC, which has no statutory authority to conduct any further review of accuracy, beyond making sure all jurisdictions submitted results.
Several other places found miscounts when they checked their presidential vote totals with hand counts.
We cannot expect that every single vote will be accurately counted. A few votes will always be ambiguously marked, a few ballots processed incorrectly for one reason or another. That's why Wisconsin statutes make it pretty easy for candidates who lose elections by less than 0.25% of the votes to get recounts. We don't want winners chosen by small, random errors.
However, miscounts larger than 0.25% are not supposed to happen. That's why Wisconsin statutes make candidates who lose elections by more than that pay for the recount themselves if they want one. We don't want taxpayers to have to finance recounts in which the winner won't change when the random errors are fixed.
Despite the accuracy standard in the state's recount law, Wisconsin's 2016 recount found error rates in many jurisdictions larger than 0.25%. The most frequent were human errors when tabulating write-in votes. Inaccurate write-in totals were found in nearly every county. Statewide on Election Night, election inspectors had missed 15.6%, or more than 1 in every 7, of the votes for Evan McMullin, the most popular write-in candidate.
But worrisome electronic miscounts were found, too, particularly with one model of voting machine, the Optech Eagle.
For their Election Day count, the City of Marinette had used three Optech Eagles to process their mailed-in, early, and absentee ballots. Marinette County conducted its recount manually and discovered that the three machines had missed a significant number of valid votes. One voting machine failed to detect votes on 30.8% of the ballots. The undervote rate had been obvious on the poll tapes printed out on Election Night, and recount minutes note that the poll workers had noted problems with the machines.
However, both the municipal and the county canvass originally certified the miscounted results, and did not correct them until forced to do so by the recount.
City of Marinette Ballots counted by Optech Eagle, November 8, 2016
Election-night opscan poll tapes,
Wards 7, 8
Number of ballots counted by the Optech Eagle 1
Undervotes reported on the poll tape1
Percent of ballots read as blank by the voting machine
1 – From the poll tapes, which were included in the Marinette County recount minutes.
2 – The number appears to be 39 on the poll tape photocopied in the recount minutes, but the text of the recount minutes says 29. If the correct figure is 39, the blank-ballot rate is 12.9%, or about 1 in every 8 ballots.
Marinette and WEC officials attributed the miscount to the inability of the Optech Eagle to count any votes marked in ink not containing carbon. However, this does not explain why the miscount rate varied so greatly among the three precincts. Absentee ballots would have been marked either at a single location (the municipal clerk’s office) or at home, providing no apparent reason why voters in one precinct would have used incorrect pens at more than three times the rate of voters in another.
Following the recount, in January 2017, WEC ordered a few manual voting-machine audits in randomly selected municipalities where counties had used computers in the recount. Three Optech Eagle machines were selected.
One of these was the Village of Hortonville in Outagamie County. When 1,051 ballots were counted by hand, after having been counted on Election Day and in the recount by Optech Eagles, 15 previously uncounted votes were found, revealing an error rate of 1.42%, or 1 in every 70 votes. Both city and county officials attempted to discover the cause of this miscount, as did WEC staff. None could clearly ascertain the cause of the miscount, and WEC staff concluded “This exercise did not produce a result that allowed staff to understand how the Optech Eagle treated these ballots with confidence,” and “The analysis of the performance of the Optech Eagle identified a significant limitation of the equipment.”
A second manual audit of an Optech Eagle, conducted by the City of Eau Claire, found an error rate of 0.35%, according to material submitted by the City to the WEC. While less serious than the other miscounts, this is still larger than the 0.25% error rate anticipated in statute, and so should be considered unacceptable.
WEC has issued a strong warning to jurisdictions that continue to use the Optech Eagle. Most miscounts that could be diagnosed were caused by the type of ink used to mark the ballots. As a result, the WEC is strongly recommending that election officials recopy all mailed-in ballots onto new ballots, and cast the copied ballots rather than the originals.
Signs that Racine County certified incorrect Presidential vote totals.
Racine County used Optech Eagles to count more than 98% of its ballots on both Election Day and in the recount. If Racine County's Optech Eagles had miscounted on Election Day for the reason the Wisconsin Election Commission believes they miscounted elsewhere--that is, an inability to read the ink in which many mailed-in ballots were marked--Racine's votes would have been miscounted both times.
Therefore, it will take a manual count to determine whether votes were counted correctly. Only one Optech Eagle used in Racine County has been checked with a hand count. The Village of Mount Pleasant in Racine County was the third Optech Eagle selected by WEC for a manual voting machine audit. When 1,877 ballots were manually counted, 16 previously uncounted votes were found, for an error rate of 0.83%.
But what about those Racine results that no one has yet checked for accuracy?
Voters are not required to vote in every race on the ballot. In the 2016 election, about one in every 130 Wisconsin voters who cast ballots, or 0.77%, chose not to vote for president.
Undervote rates--the percentage of ballots on which the voting machines detected no presidential vote--much higher than the overall statewide rate could be a sign of miscounts of the same type discovered in Marinette, Hortonville, Eau Claire, and Mount Pleasant.
And Racine County's certified results do contain such a sign. In fact, results from 61 of Racine County's 68 polling places show undervote rates higher than the statewide rate.
Countywide, the Board of Canvass certified election results that included no presidential vote for 1.78% of the ballots countywide—more than twice the statewide undervote rate.
Within the City of Racine, certified results indicate no presidential vote was counted from 2.6% of the ballots. This result can be true only if 1 in every 38 City of Racine voters chose not to vote for president.
Results for six precincts are even higher—more than five times the statewide blank-ballot rate, up to one ballot in every 12 being counted without a vote for president.
Racine County Reporting Units with recount-certified results
indicating blank-ballot rates more than 5 times the statewide rate
Municipality / Ward
Percentage of ballots from which no presidential vote was counted
City of Racine, Ward 26
Village of Elmwood Park, Ward 1
City of Racine, Ward 35
City of Racine, Ward 1
City of Racine, Ward 5
City of Racine, Ward 25
Finally, there is some direct-observation evidence that Racine County's voting machines were miscounting. During the recount, observers were able to see each ballot well enough to count votes as the ballots were fed into the voting machines. In several precincts, the observers saw more votes than the machines counted. Their observations were consistent with the types of miscounts now documented in other jurisdictions.
However, when the observers reported these findings to the county officials, county officials refused to check the machines' accuracy, even in a precinct with fewer than 350 ballots. The request for a hand count to verify accuracy was energetically refused, as a deputy county clerk declared "I don't care if it's only five (ballots). I am not going to do a hand count for anybody."
Voter confidence, and our right to self-government itself, demands accurate vote-counting.
Neither voters nor election officials should shrug off such clear and serious signs of miscounted election results.
Even if they did not change the outcome this time, miscounts in excess of the 0.25% error rate written into in Wisconsin's recount law are dangerous. If we allow miscounts larger than that to remain undetected and uncorrected, we create a situation in which election results can be both wrong--if an error rate of more than 0.25% changed the outcome--and not subject to recount unless the 'losing' candidate can raise a lot of cash very quickly.
To sustain voter confidence, our election officials must adopt a responsible, managerial attitude toward accuracy.
They cannot continue to be unwilling to resolve well-founded suspicions of miscounted votes.
What is an 'undervote'?
Voters are not required to vote on every race on the ballot. They can vote in some races and leave others blank. Typically, nearly every voter will vote in the race at the top of the ballot—last November, that was the President. Farther down the ballot, they tend to leave more races blank.
When a voter skips over a race, and casts no vote for any candidate who in on the ballot or any write-in, that is called an 'undervote.'
Finer detail: It is also an undervote if the voter votes for more candidates than is allowed in the race, and declines to fix the problem if the ballot is rejected. An undervote should also be registered if the voter marks the ballot in a way that neither the voting machine nor an election official can tell the voter's intent.
Why is the undervote rate important?
A high undervote rate doesn't prove that the votes were miscounted, but it is a red flag that indicates the need for verification.
Most of the more serious, known miscounts in Wisconsin elections and in other states were apparent in a high undervote rate.
Some, like the miscounted Stoughton referendum of 2014, showed up as dramatic undervote rates. In that race, a programming error caused the machines to read none of the valid votes, making it look like 99.7% of the voters had ignored the referendum. (A separate malfunction caused one machine to count votes that were not really there, which is what provided that extra 0.3%.)
Others miscounts can reveal themselves in undervote rates that are not as dramatic, but still suspicious. For example, electronically tabulated vote totals from one voting machine in the City of Marinette last November 8 showed that it had counted no vote from 128 of 416 ballots--an undervote rate of 30.8%. Sure enough, this was discovered to have been an electronic miscount when examined in the recount.
Elsewhere, machine malfunctions such as overheating caused the voting machines to miss some of the votes.
Deliberate manipulation might not show up as a high undervote rate--at least in counties where the officials look at undervote rates. More likely, the dishonest programmer would 'flip' votes--that is, have the machine add some of one candidate's votes to the other's total--so the fraudulent totals would not look suspicious.
Finer detail: Not every high undervote rate will turn out to be an error. For example, one ward in the City of Madison tends regularly to produce a higher undervote rate than other wards. In this ward, a large number of voters are nursing home residents who vote as 'permanent absentee' voters. Ballots are routinely mailed to them, without them having to submit a new request for every election--but only as long as they return the ballots. If they have no preference or interest in any of the races on the ballot, they will return an entirely blank ballot, just to maintain their status as a permanent absentee voter.
How do you calculate an undervote rate?
To calculate the undervote rate for any race, add up the votes counted for all the candidates in that race.
Subtract that number from the number of ballots cast. That gives you the number of ballots from which no vote was counted.
Divide the number of blank ballots by the total ballots, and you have the undervote rate.
This link will give you the total number of votes cast and counted in last November's presidential election, by ward, in Wisconsin.
This link will give you the total number of ballots.
That sounds easy. Why do I need any more instruction than that?
It should be easy, but when calculating undervote rates in Wisconsin, we have a special challenge.
Believe it or not, in many counties, including Racine, the election officials don’t count all the votes—deliberately. We don't really know how many total votes there were.
The uncounted votes were write-in votes for protest candidates, like Bernie Sanders. Wisconsin law was amended in 2015 to remove the requirement that election officials count protest votes. They are required to count only the votes for candidates who are on the ballot and write-in candidates who have registered. Counties that want to continue to provide full election results have the option of counting and reporting the protest votes--and some do. Our election officials call these votes--somewhat dismissively--'scattering' votes.
Setting aside the question of whether those votes should be counted, the fact is that in last November's election only some Wisconsin counties did count them. Based on the figures reported by the counties that reported all the votes after the recount, we can estimate that almost one in every hundred Wisconsin voters--0.955% to be precise--cast a scattering vote for president.
If you are working with the election results from any county that did not count or report scattering votes (Racine County did not), you need to increase the total number of votes by a number equal to 0.955% of the total ballots. This will be an as-good-as-you-can-get estimate of the actual number of votes for president.
What was the statewide undervote rate in last November's presidential election?
When adjustments are made to estimate the number of scattering votes for those counties that did not count them, we can calculate that 0.77% (about three-quarters of one percent) of Wisconsin voters, or about 1 in every 130 voters, cast a ballot without a vote for president.
Undervote rates significantly higher than that, in any individual jurisdiction, should cause election officials to look more closely to make sure that all the valid votes were counted.
Here's why we think more than 1,000 Racine County voters were disenfranchised last November in Racine.
“If an Elmwood Park poll worker had been grabbing the ballot from every twentieth voter and ripping it up, while county election officials looked on and did nothing, this result would have been pretty much the same.”
“The poll workers were not throwing out votes,” Racine voter Scott Farnsworth explained. “The problem is that county canvass officials have no process in place to notice or correct predictable electronic miscounts.”
Farnsworth participated in a hand count of ballots from last November’s presidential election, held at the Racine County Courthouse on November 14 and 15. County Clerk Wendy Christensen and county staff displayed the ballots to about two dozen volunteer vote-counters in response to Farnsworth’s open-records request.
The audit confirmed the group's suspicions: More than 1,000 Racine County voters may have been silently disenfranchised because of election officials' failure to check the accuracy of the computer-generated vote totals.
The Citizens' Audit was necessary because on Election Day, the voting machines indicated a weirdly high percentage of ballots contained no presidential vote. This is called the 'undervote' rate. But within two weeks, the County Board of Canvass had certified those totals as final--without checking their accuracy.
“A few voters always decide to leave the race blank,” explained Karen McKim, coordinator of the statewide group, “but not one in twenty. Anyone could see that those results were flawed before looking at even one ballot.”
Yet a few weeks later, in the Wisconsin recount, the county officials did not actually recount the votes. They merely ran the ballots back through the machines, which had not been reprogrammed. Observers noticed the machines were not counting all the votes, and again the machines' totals indicated suspiciously high undervote rates. But again the county officials certified the flawed results, again without checking their accuracy.
Organized by Wisconsin Election Integrity and financed by a successful GoFundMe campaign, the citizens' audit counted votes from six wards.
The audit found that county-certified vote totals had missed 2.5% of the valid presidential votes across all six wards--1 in every 40 votes.
The highest error rate was in the City of Racine’s Ward 26, where election officials failed to count 6.1% of the votes, even during the recount. More than 1 in every 17 voters were disenfranchised in that ward. Detailed results are here, for each candidate and ward.
Wisconsin Elections Commission officials believe the voting machines failed to detect the votes because voters had marked ballots with types of ink that the machines could not detect. After other counties hand-counted the recount and discovered the high rates of missed votes, the WEC decertified the machines (prohibited their future use in Wisconsin) in late September.
County Clerk Wendy Christensen has not yet publicly explained why the county board of canvassers chose twice to certify the results as 'correct and true' without checking, despite the obviously suspicious number of missing votes.
"We needed the manual count to get the truth,” said Village of Pleasant Prairie voter Liz Whitlock, who was among the recount observers who could see the voting machines missing votes.
Racine municipalities are in the process of replacing the unreliable voting machines. However, new machines do not eliminate the need for routine accuracy checks. Any computer, including both new and old voting machines, is continuously at risk of producing flawed output. Threats include both sophisticated international hackers and more mundane problems such as human programming error and random computer malfunction.
“Other local government officials make a habit of checking their computers’ accuracy,” said Whitlock. “You don’t see citizen volunteers having to audit computer-tabulated property tax bills or county park receipts. Election officials need to accept similar routine responsibility for the accuracy of our computer-tabulated vote totals.”
What needs to be done--before November 2018!
Before the audit, Farnsworth had said, “The 2016 election is over and done. This is about our future elections.”
In Racine County, as in other Wisconsin counties, a county board of canvass bears responsibility for the accuracy of results in statewide races such as president. That board consists of the elected county clerk and an appointed representative from each of the two major parties.
Current state law leaves it to these county boards to select the procedures they will use to review and approve preliminary election results. Wisconsin's county clerks could, without any legislative action, start immediately to check accuracy.
At a minimum, they should follow the Wisconsin Election Commission's written advice to check to see whether "there is a large difference between the total number of voters and the votes cast" in the top-of-the-ballot race. WEC has already warned the county election officials that "a large drop off between these two numbers might signal a problem with the voting equipment." Because WEC has no oversight responsibility for the elected county clerks, citizen observation at canvass meetings is critical to ensure that county officials begin to follow basic instructions.
But Wisconsin candidates and voters deserve more than the minimum that our county officials can get away with.
In Wisconsin and elsewhere, local election officials cannot effectively prevent electronic miscounts, including hacking. Lacking IT expertise and authority to maintain full control of the software, they must rely on others—primarily voting machine vendors and service technicians—to maintain security.
But in states that use paper ballots—including Wisconsin—local election officials can detect and correct any miscounted preliminary election results, even if they cannot fully protect the voting machines or their software. By checking the vote totals against the paper ballots, they can prevent any miscounts or fraud from ruining an election.
And that is just what election-administration authorities recommend they do. In August 2017, the US Elections Assistance Commission wrote, “Carefully conducted post-election audits mitigate error and check the accuracy of election results. Comprehensive and transparent post-election audits raise the level of public confidence in the electoral process.”
Outside Wisconsin, other states are moving rapidly ahead to implement effective election-verification practices. Wisconsin’s county clerks could implement more prudent canvass practices now, with no change in state law.
Wisconsin Election Integrity urges Wisconsin county clerks to adopt modern effective audit methods to catch and correct fraudulent or incorrect election results.
We encourage every Wisconsin voter to contact his or her municipal and county clerk to demand verification before election results are declared final.
Additional relevant information:
- Wisconsin’s biennial voting machine audits, mandated by statute, are completed only after the identified winners have been sworn into office and do not include enough machines to verify any statewide outcomes. Those audits are designed to audit voting equipment, not election results.
- Pre-election voting machine tests cannot prevent Election-Day miscounts, particularly those caused by electronic manipulation, which would be designed to flip or ignore votes only on Election Day.
- The 2016 recount did not verify voting-machine accuracy for about half the ballots in the state. All the largest counties except Dane—including Milwaukee, Waukesha, Brown, Walworth, Washington, Rock, Racine, and Kenosha—‘recounted’ by running the ballots back through the voting machines. Even with only half the ballots actually recounted, county officials changed at least 17,681 votes between the results they originally certified as ‘correct and true’ and the results they certified after the recount.
Here are links to more information about:
- How we calculated Racine County’s undervote rates (the election officials don’t, but they do release election statistics that allow the public to);
- The evidence indicating that Racine County’s voting machines miscounted both on Election Day and in the recount;
- A quick description of Wisconsin’s processes and procedures for counting votes and turning preliminary election results into final election results; and
- A discussion of standards for accurate vote counting, and why we need to watch for miscounts even when they don’t change the outcome of the election.
- Finally, for an enlightening discussion of the merits of hand-counting versus machine counting, with nationally renowned experts discussing Wisconsin's elections, read this edited transcript of a hearing in Dane County Court on November 29, 2016. The Judge decided that state law prevented her from ordering counties to hand count, but the Judge also made it clear that: 1) county boards of canvass may choose to hand count; and 2) all the evidence supported hand counts as the only reliable way to verify computer-tabulated election results.
Our hand count documented: In the 6 audited wards,
1 in every 40 presidential votes remained uncounted
as county officials declared results final.
Results for individual wards and candidates are here.