This is the fourth article in our series examining election integrity. 

The term black box voting was popularized by Bev Harris, owner of the website and author of the 2004 book by the same name.  The book defines black box voting as Any voting system in which the mechanisms for recording and/or tabulating the vote are hidden from the voter, and/or the mechanism lacks a tangible record of the vote cast.

You can watch the Emmy-nominated HBO film Hacking Democracy as it follows Bev Harris and her band of citizen activists while they show how America’s votes can be stolen without a trace.  The voting machines in the film are the same ones that counted the votes in the 2016 presidential election.

Although problems with black box voting equipment have been reported after every major election since 2000, many people became aware of it most recently when Green Party presidential candidate Jill Stein attempted to recount votes in Michigan, Pennsylvania, and Wisconsin.  Although there were problems in many states, these three swing states handed the Electoral College to Donald Trump by razor-thin margins. Michigan’s margin of victory was 0.3%; Wisconsin’s was 0.7%, and Pennsylvania’s was 1.2%.  In all three states, Clinton had been  ahead in the polls.

A group of computer scientists and election lawyers urged Clinton to call for a recount in all three states because of evidence suggesting vote manipulation or hacking. The group included voting-rights attorney John Bonifaz and computer scientist J. Alex Halderman, director of the University of Michigan Center for Computer Security and Society.  In addition, UC Berkeley statistician Phillip Stark and MIT cryptographer Ron Rivest, both advisors on the U.S. Election Assistance Commission, recommended conducting a “risk-limiting” audit.  Experts agreed that the election results could have been tipped by manipulating the vote count in a small number of jurisdictions in battleground states.

There were several reasons cited for the suspicions about manipulation.  One was the fact that the exit polls had predicted that Clinton would win, especially in swing states Ohio, North Carolina, Pennsylvania, and Wisconsin. Dr. Jonathon Simon, cofounder and executive director of Election Defense Alliance and author of CODE RED: Computerized Election Theft and The New American Century, says:

“There are huge disparities, way outside the margin of error, pretty much all in the same direction and, yeah, outcome reversing. So the question becomes, what do you believe? I’d be the first to say we’re not going to prove anything with exit polls. But what we really have is two sets of really lousy evidence. We have exit polls — you could ask, ‘Why would you believe them?’ — and you have the vote counts. And you could also ask, ‘Why would you believe them?’ They’re concealed, they’re computerized, they’re outsourced, they’re privatized — whoever is programming them has the basic control over how those votes are counted or how fictitious it could be. They’re also subject to outsider hacking, as well as insider rigging.”

A second reason to question the integrity of the election results was a suspiciously high rate of undervotes for president, which means that voters filled out a ballot, but left the vote for president blank. Finally, there was the fact that Clinton received more votes in counties where voters used paper ballots compared to counties that used fully electronic voting.

These questions led experts to warn about the problems with the accuracy of electronic voting machines.  (Voter suppression also played a big role, but we will address that in a future blog.) Although the Democrats made a lot of noise about Russian hacking, the fact is that it would be much easier for Republicans than for Russians to manipulate the election.  (There was also a conspiracy claim made by Republicans: They claimed that voting machines were owned by left-leaning George Soros, but Politifact rated this claim a Pants on Fire.)

Reportedly, the Hillary Clinton campaign and the Democratic National Committee (DNC) didn’t want to be associated with a recount at first because they did not find enough evidence to warrant one, and because they didn’t want to be accused of trying to overturn the election.  However, after Jill Stein had petitioned for a recount and raised millions, the Clinton campaign announced that it would participate. They noted that many states, including Michigan, do not automatically audit election results “to ensure accuracy and public confidence in the election.”


In Michigan, Stein requested a recount of the nearly 4.8 million ballots cast because the number of blank ballots were a “red flag” to her.  84.000 Michigan ballots had no presidential choice, which would suggest that those ballots were not counted, either because of error or intention.  Trump won Michigan by a margin of only 10,704 votes over Clinton (47.5% to 47.3%).

Michigan does not have early voting, so all the problems reported occurred on election day.  Pamela Smith, president of Verified Voting, received so many calls from the Detroit area that she reported them to state officials.   “They were running out of storage space for the ballots,” she says, noting that paper ballots were piling up on desks and other unsecured surfaces as technicians raced to fix the machines. “If you’ve only got a partial count from a voting place, something has to happen to the rest of the ballots.”  In Detroit,74.020% of voters are registered Democrats vs. 24.620% registered Republicans, which would suggest that 74% of those uncounted ballots would have voted Democratic.

Nevertheless, a federal judge halted the Michigan recount before it could be completed. The rationale given was that Jill Stein was not an “aggrieved party” because she could not have won the election. Of course, Stein and her campaign did not expect to win; they had always been very clear that they were acting to make sure the election results were authentic, not because they thought she had won.  Several states allow any candidate who was on the ballot to request a recount. Stein appealed the judge’s decision, but the Michigan Supreme Court declined to hear the appeal.

However, before the recount was stopped it did reveal that 87 optical scan machines had broken down in Detroit on Election Day, resulting in ballot discrepancies in 59% of the precincts.  Detroit elections director Daniel Baxter said that the city has had the same problem with the equipment in nearly every election.  The optical scanners were designed to register and count the votes that were marked on paper ballots, but they did this in an inconsistent fashion.  The good news is that there were paper ballots that could be counted as a backup.  The bad news is that, even if the judge had not stopped the recount, the precincts in question were barred from recounts by Michigan law. When there is a discrepancy between the number of votes counted in the electronic poll book and those actually in the ballot box, that precinct is barred from recounting.  Michigan voting laws also specify that votes cannot be recounted if there are any mistakes made by poll workers.  I don’t know what the rationale was at the time the law was passed, but the effect is to prevent recounts in the precincts where they are needed.

Without the ability to audit elections for accuracy, there is really no basis for Michigan to improve its election procedures.  “By doing audits you actually increase voter confidence,” says Smith of Verified Voting. “And that’s not gonna happen.”

After the election there were many confusing reports of overcounts in some precincts as well as undercounts, so to allay voters’ concerns, the state of Michigan did its own recount.  However, the state was also barred from recounting the precincts that had the problems.  The following is quoted from The Detroit News:

Overall, state records show 10.6 percent of the precincts in the 22 counties that began the retabulation process couldn’t be recounted because of state law that bars recounts for unbalanced precincts or ones with broken seals.


The problems were the worst in Detroit, where discrepancies meant officials couldn’t recount votes in 392 precincts, or nearly 60 percent. And two-thirds of those precincts had too many votes.


“There’s always going to be small problems to some degree, but we didn’t expect the degree of problem we saw in Detroit. This isn’t normal,” said Krista Haroutunian, chairwoman of the Wayne County Board of Canvassers.


The Michigan Bureau of Elections audited 136 (out of 662) Detroit precincts from the Nov. 8, 2016 general election.  They found no evidence of voter fraud or widespread equipment failure that would explain the discrepancies.  They mostly blamed human error and promised to work with city officials to better train precinct workers.

Tom Barrow, who is a former Detroit mayoral candidate, said:

“The city is responsible. (City Clerk) Janice Winfrey is responsible. This didn’t happen because of crazy, dyslexic senior citizens who are working as poll workers, like they want to portray this. That’s people who are trying to deny responsibility.”

He has asserted on social media that Winfrey cost Clinton the election in Michigan

So far as Donald Trump’s claim of many people voting twice, a statewide review discovered 31 individuals who may have voted twice, and they have been referred to the Michigan Attorney General’s Office for criminal investigation.   Compare that to the many thousands whose votes were apparently not counted in precincts that vote Democratic—there is no investigation for them.

Remember that the definition of a black box is “Any voting system in which the mechanisms for recording and/or tabulating the vote are hidden from the voter, and/or the mechanism lacks a tangible record of the vote cast.” By this definition, optical scanners are not black boxes when there are back-up paper ballots that can be hand counted.  In the case of Michigan, it was a federal judge and state laws that turned the optical scanners into black boxes.


In Wisconsin, Trump beat Clinton by 22,748 votes out of 2.8 million cast.  This is more than in Michigan, but the difference was still less than 1% (47.2% to 46.5%).   In Wisconsin, voters use either optical scanners with paper ballots as in Michigan, or touch-screen voting machines, also known as Direct Recording Electronic (DRE) systems with voter-verifiable paper trails (VVPTs). Preliminary analysis by election experts showed that Clinton received 7% fewer votes in counties that used electronic voting machines instead of optical scanners with paper ballots.  This translates to about 30,000 votes, which would have been enough to change the outcome of the election.  The Marquette University Law School poll leading up to the election showed Clinton leading Trump by six percentage points.

Electronic voting machines are computers, and they do whatever they are programmed to do.  In some of Wisconsin’s precincts, voters use paper ballots, but they are read and counted by optical scanners. An optical scanner scans the marks on the paper ballots and counts the votes electronically. As we noted in Michigan, optical scanners do not always count the votes correctly, either by accident or by intention.

Other precincts use touch-screen machines (DREs) that are required to generate a “voter-verifiable paper trail” (VVPT).  The paper trail is supposed to increase security.  Even if a hacker alters the votes inside the machine, the paper is supposed to be available so that the votes can be recounted or audited.  However, although the mainstream media does an inexcusably poor job of alerting the public to this fact, computer scientists have shown consistently that computer printouts “provide no protection whatsoever against hacked machines,” according to Alex Halderman, computer science professor at the University of Michigan. “Someone who hacks the machine can cause it to print out whatever they want. My group did that with the Diebold DREs we hacked, for instance.”  Security seals are also useless.  In case of interest, you can watch a video made by the Computer Security Group at the University of California Santa Barbara (UCSB) showing how one person can easily hack an election on a DRE with a voter verified paper trail without leaving any evidence of tampering, and without disturbing the security seals.

This means that optical scanners are potentially safer than DREs, because with optical scanners we have the original ballot that can be counted.  However, the count can only be verified if the ballots are counted by hand.  If officials just run the ballots through the same optical scanners a second time, the recount is unlikely to be any more accurate than the first tally.

The Wisconsin Elections Commission agreed to a recount but declined to require that it be done by hand.  Jill Stein sued to require a hand recount, but the judge ruled that each of the 72 county clerks could decide how to carry out the recount.  In the end, 47 counties were counted by hand, 13 by a combination of hand and optical scanners, and 12 by optical scanners.  This means that almost half of Wisconsin voters were denied a hand recount, largely in communities of color where voting technology is most likely to fail.

Wisconsin law requires that jurisdictions with more than 7,500 people are required to use DREs, and some precincts use them even when they are not required. Elections in Wisconsin occur in about 1850 different municipalities in 72 counties, and each one chooses its own voting system. After the election, there was no information publicly available to determine how many voters used each technology, and there wasn’t even reliable information about how many voters turned up in each location.  Some counties used multiple technologies.  This made it nearly impossible to be sure whether the differences between votes cast on paper ballots vs. those cast on DREs were systematic enough to infer tampering.

In the end, the Wisconsin recount seems to have failed to provide any useful information.


Trump’s margin of victory in Pennsylvania was 44,292 votes, more than Michigan or Wisconsin, but still less than 1%  (48.2% vs. 47.5%), even though the state last voted for a Republican for president in 1988 and Democrats in 2016 won a series of down ballot offices. According to exit polls conducted by Edison Research, Clinton won four swing states (NC, PA, WI, and FL) that she went on to lose after the computerized vote counts.  The discrepancy between the exit polls and the final results were 5.1% in Pennsylvania, a result 1.4% higher than the margin of error.

Jill Stein’s lawsuit alleged that Pennsylvania’s elections operation is “a national disgrace. Voters are forced to use vulnerable, hackable, antiquated technology banned in other states, then rely on the kindness of machines,” the complaint reads. “There is no paper trail. Voting machines are electoral black sites: no one permits voters or candidates to examine them.”

However, a federal judge denied Jill Stein’s request for a recount.  This is probably just as well, because Pennsylvania’s votes are impossible to recount.  More than 80% of Pennsylvania voters use DREs without any paper trails.  According to NBC news, a group of citizens sued the state a decade ago because the lack of a paper trail made hand recounts impossible, but they lost the case.

Nevertheless, Pennsylvania does have a process for “recounts.”  The Pennsylvania Secretary of State’s Office published a handout detailing how a recount would work.  First, the county board must “make a record of the number of the seal upon the voting machine and the number of the protective counter or other device.” The board must “make visible the registering counters of the machine, and without unlocking the machine against voting, recanvass the vote cast on the machine.”  The board is then to “recanvass by examining the totals tape on each machine, which shall constitute the recount total.” If there is a discrepancy, “the county board of elections must unlock the voting and counting mechanism of the system and examine and test the system to determine the cause of the discrepancy,” the handout reads. “In this case, the counter must be reset at zero before it is tested, after which it must be operated at least 100 times. After the examination and test has been completed, the machine inspector must prepare a written statement detailing the result of the examination and test.”  No kidding.

So how did the United States get into this mess?  Once upon a time, we had voting on paper ballots that could be recounted by hand.  There were, no doubt, a few problems with ballot stuffing or whatever, but nothing like the magnitude of what can be accomplished electronically these days. We know that it is now possible to steal an election.  The question is, Did it happen?  To be continued.