Adam Maida

Updated on October 21 at 5:47 p.m.

George Mangeni registered to vote as soon as he became a U.S. citizen in 2015. Mangeni, who immigrated from Kenya, always makes sure to cast a ballot in the suburbs of Columbus, Ohio, where he lives.

“It’s just something you do,” he told me. “You are given an opportunity to select people who make influence over your lives, and so it’s important you have a voice.”

This spring, with COVID-19 ravaging Ohio, he decided to cast his vote in the primary elections by mail for the first time. The instructions were straightforward enough, especially for a network engineer like Mangeni: He applied for a ballot, received it, marked it up, signed it, and prepared to submit it. As he got ready to post his ballot, though, Mangeni felt a pang and took a photograph of it. “I had a sinking feeling that something would happen,” he recalled. “I was like, I hope it gets counted.” But he sent the ballot to the board of elections, and didn’t worry about it anymore.

A couple of months later, he received a call from the ACLU of Ohio, informing him that his ballot had been rejected. “I was surprised twofold,” he said. “One, how the heck did they find out? And then, why was it not counted?”

The answer, it turned out, was that Mangeni’s signature on the ballot didn’t match the one he used when he registered to vote. Ohio, like 30 other states, uses signature matching as a fraud-prevention measure. Mangeni sometimes uses different signatures, and he didn’t recall which one he had used to register. Under Ohio law, election officials are supposed to mail a notice to any voter whose ballot is rejected, giving them a chance to correct an error, but Mangeni said he never received a notification. His name did, however, go into a spreadsheet at the Ohio secretary of state’s office, which is how the ACLU found it. Mangeni agreed to become a plaintiff in a suit challenging Ohio’s signature-matching law.

“I felt cheated out of something,” Mangeni told me. Besides, “it was irritating. Some of the people that won the elections, I didn’t vote for.”

If Mangeni felt irritated, a voter who has her ballot rejected in the general election, especially in a tightly contested swing state such as Ohio, is likely to feel something closer to full-on rage. A political scientist at Carroll College, working on behalf of plaintiffs challenging Ohio’s signature-matching law, found that 97 percent of rejected signatures are likely to be authentic—or, for every invalid ballot, 32 valid ones are thrown out.

Even in normal election cycles, signature-matching requirements result in many ballots being rejected. Hundreds of thousands of such ballots were disqualified this way in 2016—almost all, presumably, cast by voters who had done everything right. Rejections disproportionately hit certain demographic groups, including elderly voters, young voters, and voters of color, that are expected to heavily favor Vice President Joe Biden this fall. As voting by mail surges across the country, many elections, including the presidential race, could hinge on a process that one expert recently described to me as “witchcraft.”

Earlier this year, voting-rights advocates, Democrats, and progressive groups began a major push to make voting by mail easier, and to persuade citizens to cast their ballot that way. In the midst of the pandemic, they thought it was the best way to ensure safety and get votes counted. The effort worked: Vote-by-mail rates this year are expected to easily set records. But the push to expand mail-in voting also introduces a number of risks. While some of these dangers, including a “blue shift” of late-counted votes and disruptions to the U.S. Postal Service, have been scrutinized, less attention has been paid to signature matching, the single biggest reason for the disqualification of mailed ballots in 2016.

“Everybody’s vote should count,” says Hannah Fried, the national campaign director at All Voting Is Local, a nonprofit that seeks to expand voting access. “If you’re an eligible voter and you voted, your ballot should not be rejected for a highly technical reason out of your control—because the signature was sloppy, or you cannot write in the same way you used to be able to write. There’s something fundamentally unfair about it.”

States use signature verification as a way to protect the sanctity of the ballot. (Some require only that an envelope be signed, or mandate a witness, a notarized signature, or a photo ID.) In-person voter fraud is vanishingly rare because a would-be fraudster has to present herself as someone else at a polling station. Vote-by-mail fraud is also very rare, but signatures are intended to add an extra layer of security: First, a voter has to attest under penalty of law that the ballot is hers. Second, the voter’s signature provides a way to make sure she’s really the person who signed the ballot.

How that is done, like the rest of the American election system, varies wildly from state to state and even from county to county. Some jurisdictions use the signature on a voter-registration form or a ballot request or a driver’s license. A software program might make the first cut, or humans might conduct the whole process. Examiners might have a single autograph with which to compare, or dozens. Many people use more than one name, and might use the wrong signature on their ballot. The singer Lady Gaga tweeted a timely reminder on Sunday: “When I sign legal documents, I repeat Stefani Germanotta over+over quietly in my head so I don’t accidentally sign as Lady Gaga,” she wrote, referring to her birth name.

“When you pay attention to this stuff, you start realizing how important things like the signature pad at the DMV is for election administration,” Nathaniel Persily, a law professor at Stanford University, says. “Absentee voting is not what you think of when you’re 16 and getting your first license.”

The training and rules for officials are all over the place too. Florida law says that ballots can be rejected because of a mismatch, but a 2019 lawsuit filed by Democrats complained that the state offered no training or procedures for officials assessing signatures, “resulting in processes that are demonstrably standardless, inconsistent, and unreliable.” Under a new law, the state must offer standardized training, and a two-hour training video is posted online. You can also watch an Oregon training session on YouTube. Colorado, which conducts its elections via mail, posts its signature-verification guide online.

Your mileage may vary, but these materials didn’t give me a great deal of confidence in the system. It’s not that election officials aren’t trying—the presentations are earnest and straightforward—but they offer fairly minimal training to the people who will decide whether someone’s vote for president gets to count. Professional forensic document examiners are typically trained for two to three years, but even the most robust training systems for election officials are more like eight hours. Some of the judgment calls depicted in the materials are obvious mismatches, but others are much fuzzier.

The New York Times recently asked a Jefferson County, Colorado, election official to identify forged signatures, and he succeeded, but that’s probably the wrong way to test the process. Fraud is exceedingly rare; the much greater danger is that legitimate ballots will be thrown out.

“At the end of the day, officials are not trained in how to conduct signature-match verification,” Kristen Clarke, the president of the Lawyers’ Committee for Civil Rights Under Law, says. “They use procedures that would not stand up in a court of law.”

One result of this patchwork of laws and practices is widely varying rates of rejection. In a recent study of “lost votes” in voting by mail in 2016, Professor Charles Stewart III of MIT found that states that conduct their elections primarily by mail have a much lower rejection rate (0.92 percent) than those that allow voters to cast ballots by mail only for limited reasons (1.8 percent). These numbers are small, but in close elections they could be consequential. The different rates seem to stem from different philosophies about examining signatures. Stewart notes that absentee voting by mail has historically been viewed as a convenience, so voters took the risk. But if states mail all voters a ballot, that argument no longer holds.

“We believe in signature verification, and when we do that verification, we’re really confirming someone’s identity,” Colorado Secretary of State Jena Griswold, a Democrat, told me. “So it’s important to do that for us in Colorado. But I will say that we believe in accessible elections. And if you have a right to vote, you should have your voice heard.”

The bias toward inclusion is sometimes a matter of policy and sometimes a matter of law. In the Oregon seminar video, for example, a trainer explains: “You’re looking for reasons to keep the signature in, to validate the signature, rather than looking for reasons to throw the signature out. That was—in the last few years that internally has been a mindset shift that we have really focused on. We’re looking for any reason to keep the signature.” In Florida, a voter’s signature can be rejected only if a majority vote of the canvassing board concludes beyond a reasonable doubt that it does not match.

Because many states have expanded access to voting by mail, and many more voters tell pollsters they intend to use it, it’s impossible to predict rejection rates this year. Will states that are new to the game maintain their old strict standards? Or will they adopt the more lenient approach of vote-by-mail states, in the service of trying to make sure as many votes are counted as possible? These answers could make a big difference to the presidential race in close states, and certainly to down-ballot races. Officials will also face a new challenge: Many people will vote absentee who have never done so in the past, which means they’re more likely to make errors in the process, including submitting shoddy signatures or failing to sign ballots or envelopes at all. According to a study of Florida’s 2016 and 2018 elections, first-time mail-in voters are nearly three times more likely to have their ballot rejected.

Regardless of the overall rejection rates, it’s a safe bet whose ballots will be rejected most: those of the youngest voters, the oldest voters, disabled voters, and voters of color. The first three of these are relatively easily explained. As schools phase handwriting instruction out of their curriculum, young people no longer learn cursive. They are less likely to have consistent, well-practiced signatures, and as a result, are less likely to have two signatures from different occasions match. Over time, their handwriting matures too. Freda Levenson, the legal director of the ACLU of Ohio, told me about a voter who had registered with a girlish signature when she was in high school. By the time she tried casting an absentee ballot in her 30s, “she in effect had to forge her own signature to make it match.” Similarly, older voters’ handwriting is sometimes in decline. Voters who suffer from illnesses such as stroke may lose the ability to sign the way they once did. But why so many voters of color see their ballots rejected is not well understood.

Theories include greater scrutiny of votes of people of color—or of voters in minority-heavy precincts—and less understanding of the system among minority voters, due to disparities in the government services provided to them. Whatever the reason, the pattern is clear. An analysis of the 2020 Florida primary by the Stanford-MIT Healthy Elections Project found that Black and Hispanic voters’ ballots were rejected at roughly double the rate white voters’ were. The same was true in Wisconsin, where more than 10,000 Black voters’ ballots were rejected in the 2020 primary. President Trump’s 2016 margin of victory in the state was fewer than 23,000 votes.

In 2018, civil-rights groups sued over ballot rejections in Gwinnett County, a suburban Atlanta county that is nearly 30 percent Black. The county rejected more than 7 percent of mail-in ballots, a staggering figure compared with the national total of 1.4 percent in 2018, but only about 3 percent of ballots from white voters were rejected, compared with 5.1 percent from Hispanic voters, 10.3 percent from Black voters, and 13.9 percent from Asian American voters. Gwinnett accounted for nearly 40 percent of all the ballots rejected in Georgia in 2018.

“In a state like Wisconsin in 2016, or even Georgia, where the rejection rate in [Democrat] Stacey Abrams’ election [for governor] was as high as it was, signature matching could make or break the states,” says Vanita Gupta, the president and CEO of the Leadership Conference on Civil and Human Rights and head of the Justice Department’s Civil Rights Division under President Barack Obama. “That’s why there’s a lot of litigation now at the front end.”

The disparities have inspired occasional demands for states to scrap signature laws altogether. In order to end a lawsuit earlier this year, the state of Pennsylvania announced that ballots cannot be thrown out solely because an official believes there is a signature mismatch. But Gupta told me that, despite the problems with the process, there are good reasons for states to have signature-matching requirements, as long as they also provide an opportunity for voters to address problems. “It’s an important security measure that can lead to greater voter confidence,” she said—all the more relevant given public concerns about the integrity of elections.

When ballot rejections are challenged in court, judges almost always side with voters. But a challenge requires a voter to be aware that his ballot has been rejected, and to either have the resources to fight against it in court or know where to turn for help. Meanwhile, postelection wrangling is likely to be the subject of extensive litigation this year. Both parties and both presidential campaigns have spent months lining up lawyers and preparing for litigation after votes are cast.

Vote totals are far more likely to be fair and accurate, then, if mismatch issues are handled early, through what is known as “curing.” Tammy Patrick, a senior adviser at the Democracy Fund and a member of the National Task Force on Election Crises, told me curing isn’t merely a way to help get votes counted, as its detractors claim—it’s also an important anti-fraud tool, because there’s no way to know whether a rejected signature is malfeasance or just messiness unless officials contact voters. During her years working as an election administrator in Maricopa County, Arizona, Patrick said she never encountered a voter whose signature had been rejected who said their name had been forged.

“They would say, ‘I had my hand in a cast,’ ‘I was writing with my left hand,’ ‘I’d recently suffered a stroke,’ ‘I was signing on top of the mailbox or my dashboard,’” she said. “Things like that confirm the legitimacy of the signature. Those are all things you learn and you know when you make that contact.”

Yet only 18 states mandate a curing process, according to an analysis by the National Conference of State Legislatures, and these too are a patchwork, with requirements as varied as mailing a notice, making a telephone call, or simply undertaking “reasonable efforts.” The deadlines for curing range from 8 p.m. on Election Day, in Montana, to 21 days after Election Day, in Washington. Several states have new procedures, including Georgia, which passed a curing law in 2019, and Arizona, where a federal judge ruled in September that voters must have until five days after Election Day to correct errors.

Voters can take a few steps to try to prevent their ballot from being rejected. It’s useful to keep track of the signature you use for official documents and keep that consistent. Reading instructions carefully is important, especially because they are not always user-friendly. It’s also wise to mail or turn in ballots as soon as possible, to leave enough time for curing in states that allow it. Perhaps the simple best step a voter can take is to vote in person, either early or on Election Day, because voter errors are much less likely to disqualify votes cast in person. Earlier this year, some experts believed that as many as seven in 10 ballots would be cast by mail, but a recent NPR survey found that just 35 percent of voters now intend to vote that way, while a Navigator poll found a six-point drop between August and October in the number of Democrats who intend to vote by mail.

Although some of the rejection rates from primaries early in 2020 were extremely troubling, Patrick told me she expected improvement in November as states that were caught off guard by the pandemic professionalize their approach. “Between the primary and the general, many states have adopted these best practices,” she said. “I’m hopeful that those changes in the process are going to help inform the voters.”

In Ohio, a federal judge ruled against the ACLU in September, concluding it was too late for the state to eliminate its signature-matching requirement for the 2020 election. Mangeni told me he agreed with the judge’s reasoning, but still hopes the requirement will be lifted in future elections. He told me he doesn’t blame his local officials for his ballot being rejected: “If I am given a bunch of signatures, and I’m under the gun to count 50,000 ballots, it’s hard. There’s a reason they have experts in court to analyze this,” he said.

The broader problem, in his view, is that requirements such as signature verification make it harder for valid voters to cast their ballot. “For a country that promotes democracy and goes around the world telling people they have the right to do stuff, for them to make it harder for me to exercise my right to vote is hypocritical,” he said.

As for November’s election, Mangeni isn’t taking any chances. He’ll be going to his local polling place to cast a vote in person.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.