Paper ballots can be manually counted in different ways–sort by candidate and then count the ballots; stack the ballots into groups of 20 and 100 and then have counters mark tally sheets as they go through the stack one-by-one; and more.
Affordable technology–a simple digital camera hooked up to a projector–can beat all these methods on each of the four attributes of a good manual-counting method.
1. Ballot security.
Ballots must not be altered by the manual count. Sorting and stacking methods require the ballots to be handled several times, by several people, and moved around tables. When ballots are projected, only one person needs to handle the ballots, only once, and can keep them on one table, in full view.
In a manual count, accuracy is established with redundant counts—two or more people must agree on each vote, reconciling any disagreement. When counters make errors in sort-and-stack or tally-sheet methods, finding and reviewing the problem ballot can take a lot of time and ballot-handling. With projected ballots, everyone sees the same vote at the same time, so ambiguous votes can be reconciled when they are first encountered.
Faster methods of manual counting help to restrain costs, because labor is the biggest cost. Quicker counting also makes the task more pleasant for both counters and observers. Projected-ballot manual counts have accurately counted votes in one race at a rate of 100 ballots every four minutes, including time to stop to compare paired counters’ totals and resolve any differences. Depending on ballot design, two races could go just as fast.
The value of a manual count depends upon how much trust it produces in candidates and voters. In traditional manual-count methods, observers cannot see ballots well enough to verify for themselves that the votes are being counted accurately and honestly. When the ballots are projected, observers see exactly what the official counters see. In addition, because projected-ballot counts require no ballot-handling by the counters, observers can be drafted on the spot as official counters–powerfully counteracting any distrust.
A tally sheet completed in full view of all counters and observers serves as a record of the manual count.
Quick summary: A recount in this week’s Supreme Court race would be a good idea for everyone involved. Verifying accuracy is a necessity if we want to protect our right to self-government from errors, glitches, and fraud. Even the seeming winner, Brian Hagedorn, would be better off if a recount removes the shadow of doubt from his legitimacy as winner. But because our legislature, egged on by our county clerks, tightened the recount law so extremely, it’s unlikely we’ll get one.
* * *
April 5, 2019 – A statewide election decided by only 6,000 votes is frustrating for everyone. Accurate or not, it indicates no clear will of the people. Chances are last week’s Supreme Court race was determined by random events. Who got tied up at work and didn’t get to the polls? Who neglected to get a valid witness signature on their early ballot?
And of course, such a close result raises fears of manipulation.
I’ve been asked about what I think happened.
First, I can say with certainty that the vote totals are incorrect. Statewide results always are, not just on Election Night but even after the county clerks have certified them. Every recount always finds miscounts.
Few people know that the 2016 recount found at least 17,681 mis-tabulated votes, or 0.58% of the total, because news media highlighted only the change in the victory margin. That’s 1 miscounted vote for every 170 cast. In 2016, the errors were random—affecting all candidates—so correcting them did not change the outcome.
Why so many miscounted votes? Lots of reasons, caused by both man and machine. Our elections are administered by a temporary workforce, without serious IT expertise. Even the county clerks don’t work full time on elections. Most workers are only lightly trained and supervised; get no more than four days’ on-the-job experience every year; and work under enormous time pressure. Only in recounts do they examine their work to find out how well they did. They would have to be superhuman not to make lots of mistakes.
But back to this specific race. Were the miscounts bad enough to have identified the wrong winner? And were they random?
I see no obvious sign of hacked voting machines. When someone manipulates Wisconsin’s election computers to alter the outcome in a statewide race, they are most likely to mis-program the software for one or two of the big counties, and they will surely put the statewide result outside the recount margin.
But the unexpected results in this Supreme Court race came from northern Wisconsin, and the statewide result is so close that Lisa Neubauer can—if she can raise huge cash quickly—get a recount. If that was computer hacking, it took impressive effort (accessing several small counties’ systems) while also being incredibly clumsy (creating results that are subject to recount.)
If a mis-tabulation (either accidental or deliberate) flipped the outcome in this Supreme Court race, my nominee for the most likely culprit is mishandled early, absentee, and mail-in ballots.
The 2016 recount found widespread errors with absentee ballots, mostly officials rejecting envelopes they should have accepted and vice-versa. But there were other mistakes, too. In Dane County alone, the recount found more than 60 uncounted absentee ballots. Neither rejected nor cast, these ballots were simply overlooked on Election Day and later, during both the municipal and county canvasses.
So we know absentee and early ballots are often miscounted by mistake, and we know those mistakes are not noticed except in a recount. We also know that in elections as anywhere else, the best place to hide fraud is where no one looks for it, and where any oddities that are noticed are written off as human error.
So messing with absentee ballots would be a good way to manipulate a relatively small number of votes.
There are many ways to interfere with absentee, mailed-in, and early ballots. As we saw in North Carolina, you can intervene between the voter and the delivery of the ballot to the municipal clerk. (That’s one ‘hack’ that cannot be detected or corrected by a typical recount.)
Local officials can—must, in fact—exercise judgment (e.g., Is this handwriting legible?) when deciding whether to cast or reject an absentee ballot. And while they are exercising that judgment, they can see the name and address of the voter. That means they can make reasonable guesses about which candidate will get the votes if they decide the ballot can be cast. Implicit, unintentional bias almost certainly shapes some decisions; the effect could be much stronger with deliberate effort. Rarely does anyone review their judgments.
To be clear, I’ve seen nothing to indicate that such manipulation was done in this election or any other Wisconsin election, but it’s a fact that it could be done. A recount would clear that up for both those who suspect and those who deny any wrongdoing. If no recount occurs, we’ll just have to keep guessing.
But we are not likely to get a recount. For the past several years, the Legislature–urged on by the Wisconsin County Clerks Association–has tightened the recount law to make it nearly impossible to get a recount in a statewide race. Had an election held in April 2015 produced a victory margin of less than 0.50%, as this one did, Neubauer could have obtained a recount at no cost.
From the 2016 recount, we now know that a 0.58% error rate is a realistic expectation. So we can see why the old law, which made it easy to get a recount if the margin was 0.50%, was sensible. If unlike our legislators, we care about protecting our right to self-government against election errors, a recount is always wise when we could easily be declaring the wrong winner by mistake.
So we will all be well-served if Lisa Neubauer can quickly raise the cash—$2 million, based on the cost of the 2016 recount—that she’ll need to get a recount. The losing candidate is the only one who can buy a recount; our laws assume voters have no standing or interest in accuracy.
But neither she nor we should underestimate the effort needed to raise that much cash that quickly. To get the cash into WEC’s bank account by the deadline, she will need to raise it within about 60 hours of when WEC receives the last county’s official results. Jill Stein had a little longer to raise the money (legislators shortened the deadline after Stein showed it was possible), but this election is not drawing the sorts of national interest that helped Stein raise that much money that fast.
Having said all that, if the result in the Supreme Court race was ‘manipulated,’ my bet would not be on an outcome-flipping miscount, but on the success of a last-minute, under-the-radar, highly targeted, dark-money effort to motivate northern and rural voters to get out and vote for Hagedorn.
But as long as we allow that conduct to be legal (and we do), we cannot call it manipulation. If we want to put an end to that, we’re going to have to do more than complain about it. We’re going to have to fix campaign-finance laws. And to do that, we voters are going to have to get a constitutional amendment.
The Wisconsin Elections Commission met today, and I stayed for most of the agenda.
One agenda item had to do with fixing the snafus that caused a voter-registration list maintenance effort in 2017 to incorrectly ‘deactivate’ thousands of validly registered voters. (You may have heard such efforts described as ‘purges,’ a relatively pejorative term that is fitting whenever voter-list maintenance is used as a voter-suppression tactic.)
Among other things, so many voters were incorrectly removed from the registration lists that poll workers for the past several elections have had to work with two sets of poll books–the regular one for unaffected voters, and a supplemental list of voters who had been struck from the rolls but who would be allowed to vote if they showed up on Election Day and attested that they had not, in fact, moved.
There are dozens of reasons, it turns out, why State of Wisconsin computers got confused about whether these voters had moved. They have to do with things like registering a vehicle with your personal name but your business address, or buying a car for your college student in La Crosse and registering it there instead of where you vote. I won’t go into all the details. If you’re curious, you can read the staff report starting on page 72 of this document.
I spend a lot of time reading about election-integrity problems in other states. That means I read about a lot of skuzzy partisan machinations.
I also spend some time talking with local election officials. That, unfortunately, exposes me to much whining, excuse-making, buck-passing and “no law says I have to” attitude.
Here’s why the WEC discussion impressed me so much that I had to come home and write this blog post.
The discussion was pure, unadulterated problem-solving, start to finish. No one was looking for a partisan angle or opportunity. Not one single commissioner or staff member was whining. No energy was wasted on self-protective defensiveness, or on denying or minimizing the problems. I heard no attempts at buck-passing, no excuses.
Unlike what I hear when I talk to many local election officials about vote tabulation, no one at WEC was pointing out that statutes require them to do the work but don’t require them to do it right. It didn’t seem to cross any Commissioner’s mind to avoid their managerial obligation to detect, analyze, and correct problems until someone passes a law forcing them to do that, and paying them extra for it.
WEC commissioners and staff were straight-up committed to discovering the extent of the problems and what caused them, and to making sure they never happen again. Commissioners asked staff for hard data on error rates, and made sure that staff are not sending any more deactivation notices until the problems are fixed. Staff, for their part, were as committed to getting past problems corrected and future problems averted as the Commissioners were.
This is what responsible election administrators look like.
I wish all voters could have seen what I saw today. And I wish some reporter would write about it when good work gets done.
In brief: Jill Stein’s post-recount organization, Voting Justice, won a victory over voting machines companies ES&S and Dominion in Wisconsin. Stein will be allowed to bring in experts to examine the software for several models of voting machines, and those experts will be able to report their findings publicly. These machines are used not just in Wisconsin, but around the nation. That is all very good.
But this victory is only a breach in the wall of secrecy, not a demolition. Several factors will limit the benefits that will flow from this victory. If we’re going to secure future elections, we will need to thank Jill Stein and the Wisconsin Elections Commission–and keep on fighting.
An otherwise well-informed voter could be forgiven for not realizing that the 2016 Presidential recount effort, led by the Green Party’s Jill Stein, is still producing valuable results.
Corporate news media see elections as horse races. Once a winner is declared and the bets paid out, journalistic curiosity dies. Editors see no story in the poor overall quality of American election administration.
So the 2016 Presidential Recount—completed in Wisconsin, aborted in Michigan and Pennsylvania, not even attempted elsewhere—got a big media yawn when it did not change in the outcome.
Here’s some news you may have missed as a result:
Since 2016, the products of the recounts have supported organizers’ efforts to improve elections administration—with impressive success. In Pennsylvania, Stein’s organizers used a recount-related lawsuit to force that state to prohibit paperless voting machines and require routine election audits after future elections.
In Wisconsin, the recount produced a mountain of before-and-after data, from every polling place, for every candidate. Academics from MIT, Harvard, and the University of Wisconsin analyzed it. Although the recount increased neither Trump’s nor Clinton’s ultimate total by more than a few hundred votes, the researchers discovered that more than 17,600 votes had originally been miscounted—or 1 in every 170.
Those stunning findings were ignored by the news media, but the Wisconsin Elections Commission (WEC) took note. I believe that study contributed to their 2018 decision to encourage county clerks to improve embarrassingly weak canvass practices. (The ‘canvass’ is the weeks-long process after Election Day, during which local election officials should review the results to make sure they are correct.)
A unique Wisconsin law requires voting machine companies to place a copy of the actual vote-tabulating software in escrow after every election. Separate from the process of approving voting machines for sale, this requirement was originally intended to protect local governments’ voting-equipment investment. In case a company went out of business, the State would have a working copy of the most recently updated software as backup to keep the machines counting.
The law also requires the State to allow candidates to inspect the escrowed software if the candidates agree “to exercise the highest degree of reasonable care to maintain the confidentiality of all proprietary information.” The software in Wisconsin voting machines is the same as that used in other states, and any inspection could have national implications.
So the Stein campaign asked to inspect the software. They and WEC reached agreement on how Stein’s experts can examine the software without having the opportunity to steal the voting machine companies’ trade secrets. (I can hear the computer-security academics scoffing: “As if anyone would want to!”)
But when the voting-machine companies saw the agreement, ES&S and Dominion didn’t want to play by those rules. The two major players in Wisconsin’s voting-machine market wanted WEC to impose a gag order on Stein’s inspectors.
WEC said “No gag order.” The companies sued.
Their demand of the court was basically: “If you’re going to let Jill Stein’s experts examine our software, prohibit them from ever telling anyone their conclusions.” In other words, the voting machine companies wanted the judge to declare that the experts’ opinions were the companies’ proprietary trade secrets.
Dane County Circuit Court Judge Stephen Elke backed the WEC and turned the voting machine companies down. Likely sensing the main reason why any company would seek such a gag order, the judge wrote:
By way of example, a nutritionist might be given access to the secret formula for Coca-cola, which is undeniably a trade secret. It would not be a disclosure of that trade secret for the nutritionist to say, “After seeing the secret formula, I can tell you that Coca-cola is unhealthy and I would never let my own children drink it.”
This is a breakthrough. To my knowledge, independent experts have never before been able to examine “the software components that were used to record and tally the votes in an election” with the approval and support of the state elections agency. That software is defined as “the vote-counting source code, the table structures, modules, program narratives, and other human-readable instructions used to count votes.”
Whatever the inspectors find, the voting-machine companies will not be able to say “But the software you saw was only for review purposes. We had fixed that problem in the operational software.” If they make that claim, they’re admitting they violated the Wisconsin law that required them to provide the State with the software that was used in the election.
And what will their inspection find? Smart money would lay odds on serious problems. You need to know only the basics about profit-seeking corporations and low-bid procurement to predict flaws and holes in a product that the manufacturers believed was eternally protected from scrutiny.
We’ll have to wait to see how well the inspectors can describe those problems without violating their pledge to maintain secrecy about the details of the coding. However, this is the best opportunity so far.
A final benefit is, I believe, this decision’s effect on efforts in other states to compel independent examination of vote-tabulating software. With this precedent, those efforts will be more likely to go forward and more likely to succeed. Wisconsin has shown that states can require voting machine companies to allow inspection of the software that will count our votes.
The wall of secrecy surrounding the “black box” voting-machine programs has been breached. But it has not been torn down. The Stein team grabbed the opportunity that presented itself and made the most of it, but the benefits have limitations.
The first limitation is that this is a one-time event. If no other candidate ever demands to inspect the software, it will never again be inspected. We don’t yet have routine quality assurance.
Second, this decision does nothing, in itself, to banish proprietary software from public elections. If we want to get any closer to open-source software and routine quality assurance inspections, we’re going to have to force this review’s eventual findings into the public’s and legislators’ awareness. We’re going to have to do that ourselves. Count on the news media to maintain their willful blindness to any news, however shocking, that lacks a Republican vs. Democrat partisan angle.
Third, while this review will be able to identify any defects or vulnerabilities in the master copy of each systems software, the inspectors won’t be able to tell whether the election was hacked.
Computer programs can be manipulated in ways that are undetectable to even expert forensic analysts. In Brave New Ballot, Johns Hopkins University computer-science professor Aviel Rubin recounted how he annually provides his graduate computer-forensics students with programming he has hacked at the level of the binary code (that is, the ones and zeroes that underlie the human-readable source code). Years go by between the times when a grad student detects the hacked code, and grad students have been able to stump him, too.
Neither voting-machine companies nor election officials will be able to provide Stein’s inspectors with the software that actually counted votes on Election Day. In a “precinct-count” state like Wisconsin, computer tabulation takes place in thousands of separate computers—at least one at every polling place. Before each election, those machines’ software is copied, recopied, and modified for the many different sets of races and candidates that appear on the jurisdictions’ different ballots. In counties that use the Dominion ImageCast Evolution system (with the exception of Fond du Lac County), the software has even been transmitted over the internet from the Colorado manufacturer to the county clerk. Therefore, to confirm with certainty that the software in every machine was compliant in even one election, inspectors would have to obtain access to the thousands of copies. No one is in a position to collect all that software in a single place, and Stein’s experts wouldn’t have time to review it even if it could be collected.
Voting machines can be made to mis-tabulate without altering the source code that Stein’s experts will review. For example, a corrupt or lazy service technician could have installed unauthorized remote-access capability in the county elections computer (as Pennsylvania clerks discovered, to their dismay, in 2014) or in individual voting machines. Later, someone could have taken control of the machines’ output (that is, our election results) simply bypassing or overriding the authorized software.
The best possible outcome of these reviews would be that the experts will find the weaknesses, report them out, and motivate legislatures nationwide to adopt laws requiring open-source software and routine independent software inspection in every local jurisdiction in every election.
But even if that happens, it will not close and lock the door against outcome-altering mis-tabulation of our votes.
The only effective protection against hacked elections remains what it has always been: Routine, manual verification of the correct winners, using paper ballots, completed before the results are declared final.
Routine detection-and-correction is the onlysecurity measure strong enough to deter hacking while also protecting final election results against undetected error and malfunction.
It is the only way surely to protect the true voice of the people.
December 19, 2018 — “As the secret ballot transformed elections in the last century,” said Joseph Hall, Chief Technologist for the Center for Democracy and Technology, “risk-limiting audits are going to transform elections in this century.”
In a few years, Americans will look back, aghast, at our current election management. We will be amazed that we ever trusted vote-tabulating computers so much that we routinely declared winners without checking results for evidence of fraud, glitches, or human error. We will consider routine verification to be an indispensable part of managing elections, just as cash-register reconciliation is now for managing the corner convenience store.
In preparation for that day, it’s time to understand: What is a “risk-limiting audit”?
First, a risk-limiting audit is not a specific set of steps or statistical calculations. Like “home-heating system,” the term describes a function, not a method. If a system heats your home, it’s a home-heating system. If it doesn’t, it’s not. A wood-burning stove is a home-heating system. Electrified tile floors are a home-heating system. Triple-pane windows and attic insulation are not.
A risk-limiting audit is any review that imposes a precise limit, such as 10%, on the risk of certifying incorrect results in the event that Election-Night results identified the wrong winner. Any review that accomplishes that is a risk-limiting audit. If it doesn’t, it’s not. For example, pre-election voting-machine tests and hand-counting to verify a few voting machines’ results are good. But even when completed correctly, they do not precisely limit the risk that officials will detect and correct any outcome-altering miscounts.
(Though it’s not part of the official definition of risk-limiting audit, I’ll mention the other side of the coin. In the event that Election-Night results identified the right winner, a risk-limiting audit does not reduce the risk that officials will certify the wrong one. That risk stays at zero.)
You might be surprised that statistical analysis is not a required feature of risk-limiting audits. A full recount uses no statistical methods and if done correctly, limits the risk of declaring the wrong winner to zero. Therefore, it’s a risk-limiting audit. But full recounts require too much effort to be used routinely. So to reduce the time and effort needed to confirm elections, most types of risk-limiting audits use random sampling for selection and statistical processes for analysis.
Another feature of risk-limiting audits is manual inspection of voter-verified paper ballots. Until some as-yet-uninvented technology comes along, we can verify the computers’ verdicts only by comparing them against something that didn’t come from a computer. That is, we need direct human observation of the paper ballots that were marked, or at least verified, by the voters themselves.
Finally, the word ‘audit’ doesn’t mean what you probably think it means. Outside elections, independent auditors perform audits after auditees have completed the work. In contrast, a risk-limiting audit is completed by the responsible officials as part of their work of certifying election results. A post-certification review might provide useful information for the next election, but it cannot limit the risk that the wrong winner was certified in this one. Risk-limiting audits probably should have been called ‘risk-limiting reconciliation’ or ‘risk-limiting verification,’ but it’s too late to change that now.
In brief, the exercise uses playing cards to represent ballots containing votes for Black or Red. The cards are sorted into piles representing precincts; a sample is randomly drawn from across all participating precincts. Each card either builds or reduces confidence in the Election-Night results, until a pre-determined acceptable level of confidence is achieved—or is not. To do this exercise, you’ll need:
two decks of playing cards;
a pencil; and
a random-number generator.
Note: the statistics in this exercise are realistic but not precise; they are for illustration purposes only. A real election audit would use a sample size and confidence level calculated for the results being audited.
Example #1: When Election-Night results are correct
The first illustration will show how a risk-limiting audit plays out when Election-Night results identified the correct winner.
1. Get two decks of playing cards. They don’t need to be the same design, but the same shape and size will make them easier to work with. From one deck, remove the jokers and set the hearts aside. This leaves 39 cards in this deck—26 black ‘votes’ and 13 red.
2. From the other deck, remove the jokers and set aside both the hearts and diamonds. This leaves 26 cards in this deck, all black.
3. Shuffle all 65 cards together, but not thoroughly. Actual ballots will not be thoroughly shuffled; your cards don’t need to be, either.
4. Separate the cards into six stacks to represent precincts. You don’t need to make them the same size. In the real world, some precincts have more voters than others. Label the stacks “Precinct A,” “Precinct B,” and so on up to “Precinct F.”
5. Count the cards in each precinct consecutively and write the numbers on the labels. For example, if Precinct A has 12 cards, you will write “1-12” on that label. Then start Precinct B’s count with 13, so that Precinct B’s label will be something like “13-23.” Precinct C will be something like “24-35,” and so forth. If you’ve counted correctly, the last card in Precinct F will be 65. Some vocabulary: A list of precincts indicating the number of ballots in each (e.g., Precinct D has 13 ballots) is a “ballot manifest.” When you assign a unique number to each ballot (e.g., Precinct D contains the 36th through the 48th ballot), it becomes a “ballot look-up table.”
Now, imagine you’re the official who is responsible for certifying this election. You know the possibility of an outcome-altering Election-Night miscount is remote. Nevertheless, you want to: 1) deter fraud in future elections, 2) demonstrate to the voters that local elections are secure against even remote risks; and 3) make sure you don’t miss even a remote possibility of declaring the wrong winner.
So you’ve decided to give yourself at least a 90% chance of detecting any outcome-altering miscount before you declare the official winner. In other words, you have decided to impose a 10% limit on the risk that you will fail to notice and correct any such miscount. (You could choose a different level of risk, if you wish.) If the machines identified the correct winner, there is a 0% chance an audit will reverse that.
Initial Election-Night results indicated that Black got 80% of the vote and Red got 20%. While you haven’t looked at any ballots yet, you know how many were cast. Using that information, you consult with a statistician or the risk-limiting audit website to find out how big your manually counted sample needs to be to confirm the right winner or to detect the wrong one.
In a real risk-limiting audit, you would be told a specific number of ballots to draw for the first sample—up to a few hundred, depending on the margin of victory and the number of ballots cast in the race. Your statistician or the RLA website could also generate random numbers for you, to determine which ballots to draw for the sample. For this demonstration, let’s imagine you were told to select 10 ballots.
6. Create a score sheet with columns for the random number, the color of the card, a confidence score for each card, and a running confidence total.
8. Then find each card. For example, if the first randomly selected ballot was #38, you would check the ballot look-up table and see that card #38 is the third card in precinct D. Look at that card, record its color in the second column, and replace it.
If the sampled card was black, it builds confidence that the preliminary results were correct when they identified Black as the winner. Note +5 confidence points for that card in the third column, and add 5 points to the confidence total in the fourth column.
If the sampled card was red, it reduces confidence in the preliminary results. Note –10 confidence points in the third column, and subtract 10 points from the confidence total in the fourth column.
9. When you’ve recorded the color of each card in the sample, look at your total confidence score. If it is 25 or higher, you have statistically confirmed, with 90% confidence or more, that no Election-Day miscount identified the wrong winner. You can end the audit and certify the results.
The photo below shows that this audit could have stopped after the eighth card, because the confidence score reached 25 at that point. On average, an audit like this would need to inspect 14 cards to reach a confidence level of 25—if, that is, the Election Night result was correct.
If your confidence score is less than 25, select another random ballot, and another, until the total confidence level reaches 25. (Or until you realize that you messed up the first two steps of this exercise and are working with a deck in which there are actually more red than black cards.)
#2: When Election-Night results are incorrect
The second part of this exercise shows what happens when Election-Night results were wrong.
1. Retrieve the red cards you set aside, so that you are using all 104 cards from both decks. Shuffle them together. Again, this does not need to be a thorough shuffle. Some ‘precincts’ can be mostly red and some mostly black.
2. Separate the cards into 9 stacks of differing sizes to represent precincts. As in step 5 above, count the cards in each stack to create a ballot look-up table.
3. In this exercise, we know the election was a tie. But in a real election, we would not know that, because the computers told us that Black won, and we have not yet inspected any ballots. So we would give the same information to the statistician or RLA website that we did in the previous example—that is, an 80% victory for Black. Given that situation, the instructions we receive would be the same: Select a random sample of 10 ballots by generating 10 random numbers—this time, between 1 and 104, because we have more ‘ballots.’
4. As in the first exercise, use the ballot look-up table to find each card selected for the sample. Note the color of each and confidence points on the tally sheet. Check to see whether the total confidence score reached 25. In this case, it did not. It is possible that your first sample reached a confidence score of 25 or more. If so, your audit incorrectly confirmed a miscounted election. This can happen—statistical confidence is not the same as rock-solid certainty. A 90% confidence target means that 10% of the time it will be wrong. To calm your nerves, think of this from the point of view of election thieves who see a 90% chance that their handiwork will be noticed and corrected.
5. If the total confidence score is less than 25, you have not yet confirmed the Election-Night results, so you will need to expand the sample by inspecting more random ballots.
When the Election Night results are wrong, the more ballots you sample, the lower your confidence level will sink. As shown in the photo, as more and more ballots are inspected, it becomes more and more apparent that the preliminary results are just not right.
Election officials who conduct risk-limiting audits of real elections adopt written policies that address this possibility. A sensible policy, for example, might dictate that the audit will stop if it fails to confirm the outcome with two additional samples, and the effort will instead be put into a manual count of 100% of the ballots.
About sample size, statistical confidence, and emotional confidence
One concept should now be clear: A random sample of ballots is a miniature version of the entire election. The same winner will emerge from both–if both are accurately counted.
In the first exercise above, Black had more votes in the whole set of ballots from which the sample was drawn. As a result, more of the randomly selected ballots contained votes for Black than for Red. In the second exercise, Black did NOT have more votes than Red, and so we could not confirm a Black victory no matter how many ballots we randomly drew.
In other words, when preliminary election results have identified the correct winner, inspection of a relatively small number of randomly selected ballots provides strong evidence of accuracy. Conversely, if preliminary election results identified the wrong winner, inspecting a random sample of ballots will reveal the problem before officials certify the election.
Once we see that random samples reflect the true results, the next question is what size sample works best. Smaller samples reduce work, but increase uncertainty. Larger samples provide more confidence, but are more work.
This demonstration started with samples of 10 cards, which in the 65-card ‘election’ was a little more than 15% of the ballots. In a real election audit, the initial sample size would not need to be anything close to 15% of total ballots, particularly if the preliminary results indicated an outcome as decisive as this one.
To work through a real-life example, let’s look at the 2018 race for US Congress in Wisconsin’s First Congressional District. This reasonably close and hotly contested race filled the seat being vacated by former Speaker Paul Ryan. The actual results were: 325,003 total ballots; 177,490 votes for Steil; 137,507 for Bryce; and 10,006 for Yorgan.
When you plug these results along with a 10% risk limit into the online tools for ballot-polling RLAs, statistical operations built into that tool predict that you will likely be able to confirm Steil’s victory, if Steil actually won, with an initial sample of only 301 ballots. That’s only one-tenth of one percent of the 325,003 total ballots. If Steil did not truly win, auditing 301 ballots would produce results more like the second example above–forcing officials to keep expanding the sample until it was obvious that the Election-Night results were incorrect. (Instead of using +5, -10, and 25 total points as indicators of confidence, officials would have counted the votes in the sample and could have used the online tools to assess the results. See the section titled “Should more ballots be audited?” at the bottom of this page.)
If election officials did not believe that 301 ballots would provide voters with enough subjective confidence in the result (as opposed to statistical confidence), they could have selected a smaller risk limit. In this congressional election, a 5% risk limit would have needed an initial random sample of 389 ballots; a 1% risk limit, a random sample of 594 ballots. Or, election officials could adopt an audit policy that every initial sample will contain either enough ballots to support a 5% risk limit, or 1,000 ballots, whichever is greater.
Other lessons from this exercise
This exercise highlighted risk-limiting audits’ tightly focused purpose—to detect and correct any outcome-altering miscounts. This purpose is critical for election security and for voter confidence, but does not solve all problems. Election officials must perform other types of reviews to determine the cause of any miscounts and to monitor other important issues, including:
The presence of any miscounts that may have disenfranchised some voters without affecting the outcome;
The accuracy of any single precinct’s or voting machine’s tabulation;
The quality of any of the processes that determine which ballots were cast and accepted, such as issues with voter registration or ID, or whether all and only valid absentee ballots were accepted and counted.
In addition, this exercise simulated only one type of risk-limiting audit, known as a “ballot-polling audit.” Depending upon the size of the election, the type of records created by the voting-equipment, and other factors, election officials might decide to use a different type of RLA, such as a “comparison risk-limiting audit” or a Bayesian audit. Election officials do not need to read and digest the scholarly articles. Federal officials, staff in jurisdictions that have experience with auditing, and other experts are willing to help local election clerks understand the options well enough to make the right choices and get started with election auditing. A local election official can likely find useful help as close as the nearest local government official who has expertise related to auditing or quality assurance.
This exercise also probably raised some implementation questions: How can election officials draw a random sample in a election for which ballots are stored in sealed bags across hundreds of municipalities? How do you know how big your sample needs to be if you want to verify the outcomes in two or more races? Election officials who have started with risk-limiting audits have tackled these questions, and more solutions are being worked out with each new election. The solutions are not always easy or obvious, but local election officials who want to try their hand at risk-limiting audits need only to ask those with experience.
Finally, this exercise demonstrated that even risk-limiting audits might, on occasion, fail to detect miscounted election results. A 90% chance that serious fraud will be detected and corrected is the same thing as a 10% chance it will not be.
That highlights the need to keep two other facts in mind. First, a 90% chance of detecting fraud is better than the 0% chance that non-auditing election officials and their voters now tolerate. Second, the audits’ greatest value is, arguably, deterrence. When would-be election thieves contemplate a 90% risk of getting caught, there’s a good chance that election officials will have no electronic election fraud to detect.electionscounting votesWisconsinrisk-limiting auditsdemonstratio
Posted by Karen McKim · November 25, 2018 10:31 PM
No city treasurer would refuse to check the accuracy of property-tax bills. No county executive would release a report on annual expenditures without double-checking its accuracy.
Most local officials don’t need anyone to pass a law telling them: “Don’t forget to check your work!” They accept checking accuracy as a basic managerial responsibility.
But the Wisconsin County Clerks Association is officially on record: They don’t want to.
And their work product is our election results.
The WCCA statement came in response to the Wisconsin Elections Commission’s September announcement that they were considering two election-security measures.
The first proposal involved the only accuracy-checks the Commission has authority to order: audits of individual voting machines by municipal (not county) clerks. These audits are better than nothing, but have limited value. Not only are they limited to November elections in even-numbered years, but they also check only a few random voting machines, without confirming the right winners in any race.
The Commission was considering ordering more machines audited than in previous years and requiring the audits to be completed before election results are declared final.
The second proposal would move Wisconsin slightly in the direction of national election-security standards. The Commission said it was considering encouraging county clerks to perform audits of the type that if done widely, might confirm that Election-Night results had identified the right winner and give clerks the opportunity to correct the results if they had not.
WCCA’s response was swift, naïve, and irresponsible. The county clerks didn’t want the Commission to require, or even encourage, the county clerks to perform genuine election audits.
Perhaps sensing they are defending the losing side in a national trend (they are), the county clerks also described how they want to restrict this election safeguard:
They don’t wanna check accuracy until after they have certified final election results.
They don’t wanna check accuracy for any but the top race on the ballot.
And they want the State to pay extra if it even suggests they check accuracy.
I’m not making that up. The organization’s memo to the Wisconsin Elections Commission is reproduced, verbatim, below.
About delaying audits until after certification: The WCCA wrote that our paper ballots “should be treated like evidence and remain undisturbed” until after the clerks have certified the results. Join me in a prayer that the Trial Judges Association doesn’t share this idea about the proper use of evidence. Imagine our courts refusing to look at evidence until after they’ve reached their verdict.
About auditing only the top race on the ballot: The WCCA’s message could be restated: “If you force us to protect the US Senate election, we will refuse to protect the Governor’s race.” Hackers are delighted to know ahead of time which race will be off limits, and which races will be on an honor system.
About making the state pay extra for accuracy: With this statement, the WCCA clearly rejected the idea that accuracy is a normal managerial responsibility. Imagine a parks manager telling the county budget manager: “I signed off on this accounting of the user fees we collected. If you want me to make sure it’s right, you’ll need to pay extra.”
In a final Trumpian flourish, the memo’s author blatantly misrepresented the findings of a study by MIT, Harvard, and the UW Madison researchers (Learning from Recounts, 2017). The WCCA memo claimed the researchers had declared that “hand counts of election results are inherently inaccurate.” Compare that to the researchers’ actual words:
“…careful hand counting in a recount is the gold standard for assessing the true vote totals — in large part because of the greater focus on a single contest, more deliberate processing of ballots, and careful observation by campaign officials and other interested parties….”
* * *
Wisconsin statutes give the buck-stops-here responsibility for election results’ accuracy to the COUNTY clerks, and to no one else. Municipal clerks cannot verify results in federal, state, and county races; they have access to the ballots from only their own city, village, or town. The state elections agency is the legal custodian of no ballots at all; has only a few days after county certification before they must certify; and has no statutory authority to question results a county has certified.
We must insist the county clerks fulfill that responsibility. They have the paper ballots. They have the time. Modern election-audit practices would allow them to verify a few races on the ballot in just two or three days, while statutes allow them at least two weeks before they must certify the election. The only cost would be the hand-counters’ time at $10 or $12 an hour—a tiny fraction of the county’s elections-administration budget. They could randomly select just a few races for verification—just enough to deter election thieves in the races most liable to attract their interest.
And yet, collectively, they refuse.
Update: Wisely, the Commission ignored the WCCA’s whining and voted unanimously to encourage county clerks to start auditing during their canvass. And as the WCCA memo states, a few county clerks have begun voluntarily to incorporate hand-counted audits into their routine canvass procedures.
Every county clerk in Wisconsin received a memo on October 4 explaining the current nationwide move to election auditing and providing the clerks with instructions on how to get started.
Only voters, though, can make it happen. Voters who care about election security should contact their county clerk to find out whether their votes in future elections will be protected with hand-counted audits during the county canvass.
If not, the next election on February 19 will provide an excellent opportunity for your clerk to begin developing routine election-audit practices, since it will likely be a low-turnout election. Your county clerk has plenty of time before February to learn about the various methods of checking accuracy and work out his or her local procedures.
I’ve observed more than two dozen of these tests over the years. The ones I observed this week were typical. Even if you’re not an IT professional, I’ll bet you can pick out why these tests don’t protect Election-Day results from hacking—whether the hacker is an Internet cyber-crook or a corrupt voting-machine company insider.
Here, try it. Start by predicting what the hacker might try to do. First, do you think the hacker would make the malicious code miscount every single vote or only some votes?
You guessed ‘only some,’ and experts agree. When a blue-ribbon election-security task force convened by the Brennan Center for Justice worked out how a hacker would steal a statewide race in the imaginary State of Pennasota, they calculated that no hacker would likely alter more than 7.5% of the votes, or a little more than 1 in every 13. So if you want to detect hacking, your set of fake ballots—your ‘test deck’—should contain enough ballots to give each candidate at least 13 valid votes.
But Wisconsin municipal clerks typically create test decks with only one vote for each candidate—enough to catch only hacks that affect every single vote.
Second, do you suppose the hacker might instead allow the machines to count votes accurately all day, and then simply flip the candidates’ vote totals at the end of the day to give his guy the biggest total? You probably guessed yes, he might. So you would need to create a test deck that has a winner in each race, a different number of votes for each candidate.
Wisconsin municipal clerks’ pre-election test results typically contain a lot of ties–the same number of votes for each candidate in each race. Those test decks would not detect any vote-flipping hacks.
Finally, would the hacker’s malicious code kick in whenever the machine was turned on, or only on Election Day? This one is easy. Hacks would never trigger on any day other than Election Day.
This is the fatal flaw of pre-election testing as a safeguard against hacking. Hackers can program their code to trigger only when the calendar says it’s Election Day…or only when ballots are inserted at a rate typical of Election Day…or only when the machine has been operating continuously for more than eight hours…or only on some other telltale sign that real votes, not test votes, are being counted. As the Brennan Center Task Force report put it, trying to use tests like these to detect hacking would create a constantly escalating arms race between election officials trying to make the test look like a normal Election Day and hackers finding new ways to detect a test situation.
Many of Wisconsin’s pre-election tests do not hide the fact that the machines are running in test mode, not Election-Day mode. The photo at right is a close-up of the voter-verifiable paper trail from an AVC Edge voting machine, programmed by Command Central, being tested in Juneau County before the August 14, 2018 primary. Notice that the voting machine printed “PRE-LAT PAPER RECORD” at the top of the ballot. ‘LAT’ is the computer professionals’ term for “logic and accuracy testing,” a basic routine whenever software has been updated. (I don’t know why Command Central calls it “PRE-LAT”.)
This machine clearly knows it is counting test ballots, not real ones. Operating in test mode doesn’t render the test useless for things like catching innocent programming errors. But:
It is humbug for election clerks to fool themselves, or to fool the public, into thinking these pre-election tests provide any protection against hacking.
If we want to stop being fed humbug, we have to stop falling for it. If your local election officials tell you:
“Election results are protected by keeping the machines unconnected from the Internet,” tell them that you know that they have no idea about what happens to the software before it comes into their control.
“Election results are protected by federal and state certification,” tell them you know that the software has been copied and updated many times since it was certified, and that no one has ever or will ever inspect the software that will count your votes on Election Day.
“Election results are protected by the audits we already do,” tell them that audits completed only after the canvass cannot possibly protect results they have already declared final (‘certified’).
The solution: Contact your county election office. In Milwaukee County, that’s the Elections Commission; in other counties, it’s the county clerk. Tell them: “This voter is done with humbug. I know that one and only one safeguard can protect our final election results. Use our paper ballots to detect and correct any electronic miscounts before you declare election results final. Start this November.”
Don’t expect your county official to be stubborn; several are already planning to check accuracy before they certify the November results as final. Find out if yours is one.
But if your county officials are not now planning to begin auditing, don’t accept excuses. They got a memo on August 1, 2018 from the Wisconsin Elections Commission that made it clear: “A post-election audit is a tool that could be implemented to confirm that results have been tabulated accurately,” and “post-election audits of the results may be conducted prior to certification of the canvass.” The Commission even gave them basic instructions they can follow.
No more humbug about election security. Tell your county officials today: “Time’s up. Pre-certification audits. This fall.”
You can also help by donating to help Wisconsin Election Integrity get the no-humbug word out to voters, officials, and media through our 2018 publicity campaign.
Just a few tweaks to WECs’ audit policy could make Wisconsin’s November 2018 elections the most secure since we started counting our votes with computers.
July 24, 2018 — There’s still time before the November 2018 elections for the Wisconsin Elections Commission (WEC) to put a patch on the state’s biggest, most dangerous election-security hole. Up to now, local election clerks haven’t been checking the voting machines’ Election-Day accuracy before the certify election results. They could be doing that easily, quickly, and cheaply.
To audit voting machines’ November 6 output, neither WEC nor the local clerks need to spend an extra penny over what they already have budgeted for that election. The WEC has to change only one policy at their September 25 meeting.
But voters have to speak up–now. We must tell the WEC to revise their policy regarding the voting-machine audits for 2018, and order those audits to be completed in every county before election results are declared final. The WEC can be reached at (608) 266-8005 or by e-mail at email@example.com.
We have paper ballots. And local officials have up to two weeks after each election to review (‘canvass’) them to make sure the results are correct before they declare the official winners (‘certify’).
But Wisconsin election clerks seal them in bags on Election Night. During their review, they look at the poll tapes, but leave the ballots sealed. Then they certify. They swear the winners into office. Twenty-two months later, they destroy the ballots.
Perpetually sealed paper ballots do not deter hackers; they protect them.
About half the states require officials to do at least a little auditing of computer-calculated Election-Day results before they certify. But in Wisconsin, state law merely allows, but does not require, accuracy checks.
When I ask county clerks why they don’t check Election-Day accuracy, I get answers like, “If we had to count votes manually, that would defeat the purpose of using the machines,” and “If these machines were capable of miscounting, the State wouldn’t let us use them.” And “We did a recount before and didn’t find that election had been hacked.” Basically, the people who manage our voting machines don’t believe they can be hacked. Or that they can malfunction. Or that humans sometimes make programming errors.
We can’t have kind of naiveté among our voting-machine managers.
Since 2006, Wisconsin statutes have required the state elections agency to order voting-machine audits following November elections. That law, section 7.08(6) of the statutes, also orders local governments to do any audits the WEC tells them to do.
As is typical for laws like this, the statute leaves the details to the bureaucrats. How many voting machines to audit? When to audit? How to select the sample? Those decisions are left to the state elections agency.
But state elections officials have, before this year, denied the risk of an Election-Day hack. They were so confident, they didn’t think anyone needed to look for it. So they have never ordered the type of audits that would protect final election results from hackers.
But times have changed, and awareness of the complex risks–not limited to Russian hackers–has grown. WEC will be tweaking their voting-machine audit instructions soon, as they always do shortly before November general elections, and we voters have got to make sure they do it right this time.
We must demand two things.
First, the 2018 audit instructions need to tell local officials “Finish the audits during the county canvass so that you can correct any hacks or errors you might find.”
From 2006 through 2012, the State told local officials to wait to check accuracy until after they had certified the results. In 2014, state elections board members ordered their staff to stop prohibiting on-time audits. But they have never ordered timely audits—they merely stopped prohibiting them.
Second, we must demand that the WEC order audits of at least one voting machine in each county. More would be better, of course, but they’ve budgeted for only 100 voting-machine audits, and Wisconsin has 72 counties. So they can do this.
The sample selection method used in previous years is too odd to explain here. It has to do with making sure the sample contains five of each make and model of voting machine. The critical fact is that it has always left some counties out.
Wisconsin’s voting machines are, in all but a few counties, programmed at the county level. For the federal, state, and county races, the same vote-counting code is copied onto all the voting machines in a county. So there’s a good chance you could deter hackers by randomly selecting one machine in each county.
The best audit would, of course, include enough ballots to produce a statistically valid answer to “Are these the right winners?” But we’re down to the wire in 2018, and valid, respectable audits will probably need to wait until 2020. Until then, we need quick, better-than-nothing audits.
About cost: Funding for around 100 voting machine audits has already been budgeted–or should have been. Unless they increase the sample size, the WEC can order protective audits for the same price they are planning to pay for useless ones.
Just those two tweaks to WECs’ audit policy, and Wisconsin’s November 2018 elections will be the most secure in our state’s history since we started counting our votes with computers. They will be the first in which would-be hackers were put on notice: Any voting machine anywhere in the state might be randomly selected for an audit while there is still time to detect your mess and clean it up.
So: We must tell the WEC to order voting-machine audits in every county, and that they be completed before November 2018 election results are declared final.
This topic will be on their September 2018 meeting agenda, and they have asked for voters’ input.
May 31, 2018 — Ask any Wisconsin official whether our elections are secure and you’ll get this answer: “Our voting machines are never connected to the Internet; elections administration is decentralized; and the machines’ designs were approved by the federal and state governments.”
Go ahead—call your county clerk and ask. That is the answer you’ll get. He or she sincerely believes those three safeguards protect Wisconsin elections.
Here’s how many of the Wisconsin election officials’ three favorite safeguards made the list recommended by the national authorities: None.
Here’s why not:
Voting machines are and always will be programmed by fallible, corruptible humans. Anyone who develops or loads the software can manipulate the vote totals without the Internet.
You could hack into only one or two big counties’ voting machines and swing a statewide election. In fact, county control of the voting machines might make hacking easier by increasing the number of vulnerable entry points. Besides, it’s not really the local clerks who manage the machines anyway. It’s three companies: Command Central, Dominion, and the biggest one, ES&S, which supplies the software that counts about 2 of every 3 Wisconsin votes.
Federal and state approval of the voting machines’ original design cannot secure the machine that counts your votes. The software was copied and updated dozens of times before it reached the machine in your polling place. No one ever inspects or approves the software that actually counts your votes.
What does protect elections: Paper ballots and audits.
Wisconsin has paper ballots, but not a single county clerk—not one—looks at those ballots before declaring election results final.
As long as our ballots are packed up on Election Night and kept under seal until they are destroyed 22 months later, Wisconsin elections are no safer than if we were using paperless touchscreen machines.
That’s where Wisconsin’s decentralized elections can help to protect our elections. State law gives every county clerk two weeks after a general elections—and more if they request it—to review the Election-Night results to make sure they are correct. The county clerks and their boards of canvass are free to adopt any review method they choose—it’s up to them.
Any county or municipal clerk could, at any time he or she chooses, open the ballot bags and conduct a legitimate audit. It would be harder for municipal clerks, because they have only a week for review, and they certify only local races anyway. But county clerks have plenty of time to audit, and they certify the state and federal races—the ones most likely to be hacked.
Call your county clerk today and educate them. Tell them that the three safeguards they rely on are not really safeguards at all. Tell them to start—NOW—planning how they will verify the voting machines’ Election-Day accuracy after future elections. If your county clerk doesn’t know how, refer them to Securing the Nation’s Voting Machines: A Toolkit for Advocates and Election Officials, which has a good list of pointers and resources. Write to your local newspaper to make sure they cover this story.
Then, keep calling your county clerk until you get the answer Wisconsin voters deserve: “Yes, we will make sure your votes were counted correctly before we declare election results final.”
But the vote-counting system is separate. It resides on no computer that either of them can control, monitor, or inspect. It was outside their range of vision in 2016, and it’s outside their vision now. They don’t talk about voting-machine security because they don’t know.
Wisconsin’s vote-counting computers are controlled, protected, and monitored by our local election clerks and by three companies—ES&S, Dominion, and Command Central.
That’s all. No one else.
Local election clerks have exactly the level of IT sophistication you think they do. County clerks send our vote-counting software off to Nebraska or Colorado or Minnesota to be reprogrammed for each new election, with no way to notice if it comes back carrying malicious code. A few counties use an application supplied by those companies to reprogram the software themselves. They put a plastic seal on it when they’re done.
Wisconsin’s local election clerks will happily leave a service technician alone with a voting machine or the county’s election-management computer, with no way to notice if he installs malicious code or a wireless communications card. They put a plastic seal on the voting machines for Election Day.
Go ahead—ask them. They seal the software. They seal the machines. They seal the paper ballots that they could use—but don’t—to check the machines’ Election-Day accuracy.
And what about where the real danger lies—within the voting-machine companies? How well does their security guard against external hackers and corrupt insiders?
The companies themselves might not understand IT security. Professor Aviel Rubin of the Johns Hopkins University Information Security Institute checked with the major American voting-machine companies before the 2016 elections and discovered none employed “even one full-time trained expert in computer security.”
Congress, too, has been frustrated in its attempts to get straight answers about the companies’ security practices. When Congressman Ron Wyden tried to get answers from ES&S, he didn’t get a response from anyone with ‘security’ in their job title. Instead, a Senior Vice President for Governmental Relations replied, saying “At ES&S, security is the responsibility of not just one, but all who elect to work for our company.”
This governmental-relations expert reassured Rep. Wyden that ES&S had asked DHS “if they had knowledge of any such security issues involving ES&S to which they responded that they did not.” Well, whew.
ES&S—this company where every employee handles IT security and yet the vice president has to ask DHS to find out whether they’ve had a security breach—is the largest supplier of voting machines to Wisconsin. Just one of their machines—the DS200—counts more than 60% of Wisconsin’s votes, including votes from Milwaukee, Dane, Waukesha, and La Crosse Counties.
We cannot make the voting machine companies hire IT security staff before we elect a governor and a US senator in November.
And we cannot make our local election officials into IT sophisticates, ever.
But we can put an end to the honor system. That is, we can force our local election officials to use the paper ballots to detect and correct any miscounts before they declare election results final.
Our local election officials are the legal custodians of the paper ballots. They can unseal them anytime to count votes and make sure the voting machines counted right. At any time before the 2018 midterms, our local election officials could learn about results-audit practices already in use in other states and bring them to Wisconsin.
Do these three things today:
Contact WEC. Tell them to exercise leadership in getting county clerks to audit election results during the canvass. Tell them to use some of the federal HAVA funds; they’ll know what that is.
Contactyourlocalnewspapereditor. Tell him or her that you want to see local journalism take a sober look at voting-machine security right here in Wisconsin—and that doesn’t mean writing about plastic seals.