Getty / The Atlantic

As polling places closed on November 6, 2018, the expected “blue wave” looked more like a ripple. Not only had some of the highest-profile Democratic candidates lost, but the party’s gains in the House and the Senate looked smaller than anticipated.

The wave, it turned out, simply hadn’t crested yet. Over the ensuing weeks, as more ballots were counted, Democrats kept winning races—eventually netting 41 House seats. In Arizona, the Republican Martha McSally conceded the Senate race to the Democrat Kyrsten Sinema, who picked up more than 70,000 votes in post–Election Day counting. Democrats narrowed deficits in races in Florida and Georgia too. Republicans were stunned.

“California just defies logic to me,” then-Speaker Paul Ryan said in late November. “We were only down 26 seats the night of the election and three weeks later, we lost basically every contested California race.”

This sort of late-breaking Democratic vote is the new, though still underappreciated, normal in national elections. Americans have become accustomed to knowing who won our elections promptly, but there are many legitimate votes that are not counted immediately every election year. For reasons that are not totally understood by election observers, these votes tend to be heavily Democratic, leading results to tilt toward Democrats as more of them are counted, in what has become known as the “blue shift.” In most cases, the blue shift is relatively inconsequential, changing final vote counts but not results. But in others, as in 2018, it can materially change the outcome.

Although it is slowly dawning on the press and the electorate that Election Day will be more like Election Week or Election Month this year, thanks to coronavirus-related complications, the blue shift remains obscure. But the effect could be much larger and far more consequential in 2020, as Democrats embrace voting by mail more enthusiastically than Republicans. If the public isn’t prepared to wait patiently for the final results, and if politicians cynically exploit the shifting tallies to cast doubt on the integrity of the vote, the results could be catastrophic.

Imagine that as November 3, 2020, ticks away, President Donald Trump holds a small lead in one or more key states such as Pennsylvania—perhaps 10,000 or 20,000 votes—and seems to have enough states in his column to eke out an Electoral College win. Trump declares victory, taunts Joe Biden, and prepares for a second term. But the reported results on Election Night omit tens of thousands of votes, including provisional ballots and uncounted mail-in votes. Over the coming days, as those votes are counted, Trump’s lead dwindles and eventually disappears. By the end of the week or early the next, Biden emerges as the clear victor in Pennsylvania—and with that win, captures the race for the presidency.

If that’s how things unfold, Trump is unlikely to take defeat snatched from the jaws of victory graciously. He has already spent months attempting to delegitimize the election system. So imagine that he instead cries fraud and insists he’s the target of a criminal Democratic coup. What if he encourages his supporters to take to the streets, where there are violent clashes between partisans? He might even urge the Republican-led Pennsylvania General Assembly to submit a slate of Trump-backing electors, citing the Election Day returns, even if the full tally clearly shows Keystone State voters chose Biden.

The hypothetical of a blue shift reversing the early projected winner is the “nightmare scenario,” according to the election-law expert Rick Hasen. Either Trump or Biden could win by a sufficient margin to make the result clear on Election Night; it’s also possible that multiple states might see a decisive post–November 3 blue shift, creating even more chaos.

“You don’t need to worry about Russia,” Edward Foley, a law professor at Ohio State, told me. “Simply anxiety over a blue shift and willingness to litigate about it and fight about it could cause a raging contestation over a presidential election.”

The blue shift is the product of two major developments in elections over the past 70 years.  First, Americans began to expect that they would have results on Election Night itself. In the first national elections, it was impossible to gather results from many different jurisdictions promptly, and even then, there was no way to instantaneously deliver the results to the public. Electronic communications began to change that. Abraham Lincoln learned he’d won in 1860 by staking out the telegraph office until the wee hours of the morning. But when the races were close, or the votes were slow to be tallied, even instantaneous communications couldn’t deliver a result that hadn’t yet been determined. Nearly a century after Lincoln, in 1948, CBS News’s Edward Murrow signed off without being able to give the result of the close election between Harry Truman and Thomas Dewey. (The Chicago Daily Tribune was not so patient.)

The second change was the introduction of a technology that allowed television networks to project who would win the election, sometimes even before the last polling places had closed. In 1952, for the first time, CBS and NBC each experimented with using computers to analyze the early returns, and by 1960, they were a key part of the election coverage. Television became central to Americans’ Election Night rituals, and the networks’ projections came to stand in for actual results. Strictly speaking, there are no election results until boards of elections certify them. Practically, Americans usually assume that whatever the TV tells them is fact.

“From a legal perspective, there are no results on Election Night, and there never have been,” Foley told me. “The only thing that has ever existed on Election Night are projected results that the media has helpfully provided to its audiences.”

Eliding this distinction created a disaster in 2000. On Election Night, networks projected a win for Democrat Al Gore, then withdrew it, then predicted a victory for Republican George W. Bush, and then withdrew that too, as results in Florida were too close to know the outcome. A final answer wouldn’t come until December 12, when the Supreme Court slammed the door shut on a recount, ensuring that Bush’s lead in Florida would stand, and thus that he would be president.

The mess of the 2000 election in Florida—butterfly ballots, opaque instructions, hanging and pregnant chads—inspired the 2002 Help America Vote Act. Among other things, the law requires that voters who believe they are eligible to vote but who don’t appear on a voter register be allowed to cast provisional ballots that are adjudicated later. A growing number of states also adopted no-excuse absentee ballots, which were intended to make it easier for people to vote without waiting in long Election Day lines.

In 2012, while watching the Ohio returns, Foley wondered what effect votes counted after Election Day might have. He found something astonishing. Looking at five battleground states, Foley discovered that from 1960 to 2000, there’d always been some change between the Election Night tally and the final results, usually in the hundreds or thousands of votes, and sometimes favoring either party. Starting in 2004, the size of the shifts had become reliably Democratic and significantly larger—nearly 80,000 votes in Virginia in 2008. Foley christened this effect the “blue shift.”

The blue shift remains little studied and poorly understood. In a 2015 paper, Foley and the MIT political scientist Charles Stewart III found evidence that the blue shift was correlated with the number of provisional ballots cast. A California Institute of Technology paper this year studying the blue shift in Orange County, California, found that many provisional voters are younger, more likely to be nonwhite, and more transient, all populations that tend to vote for Democrats. Foley and Stewart also found no strong correlation between mail-in or absentee ballots and a blue shift. These votes have not historically had a strong partisan leaning.

“I think an honest assessment of this is that we’re still learning,” Foley said. “While we have made some progress as a field, I don’t myself feel confident that we’ve really pinned down causality.”

Meanwhile, politicians and the general public seem largely unaware of the phenomenon, which is one reason Paul Ryan was caught so off guard by the 2018 results.

“I can understand why professional politicians would be anxious if their Election Night leads are slipping away when they’re used to the expectation that you could bank on an Election Night result,” Foley said.

Ryan was quick to say he didn’t question the final California results. But not all Republicans were so scrupulous. (Democrats lodged some of their own claims about stolen elections, especially in the Georgia gubernatorial race.) Even though there was no indication of fraud in Orange County, Trump likened the slow tally of the votes there to a massive ballot-fraud operation in North Carolina, which resulted in an election do-over. And as Democratic candidates for governor and the U.S. Senate in Florida gained votes in post–Election Day counting—roughly 20,000 a piece—Trump grew agitated: “The Florida Election should be called in favor of Rick Scott and Ron DeSantis in that large numbers of new ballots showed up out of nowhere, and many ballots are missing or forged. An honest vote count is no longer possible-ballots massively infected. Must go with Election Night!” The state’s sitting senator, Marco Rubio, and then-Governor Rick Scott, who ultimately won the Senate race, both echoed that claim.

This was nonsense: The ballots that were being counted had been cast legally, and to not count them—to “go with Election Night,” in Trump’s parlance—would have constituted widespread voter disenfranchisement. In the end, the votes were counted, the Republicans won, and the fuss quieted down. But Trump’s fury offers a small taste of what might happen in a similar situation—only with Trump himself on the ballot, and the presidency at stake.

A blue shift this November is all but certain; the real question is how big it will be. In a 2019 paper on the prospect of a disputed 2020 election, Foley wrote:

It is not unreasonable to expect Trump’s Democratic opponent in 2020 to gain on Trump by over 20,000 votes in Pennsylvania during the period between Election Night and the final, official certification of the canvass. The key question is whether this kind of gain simply extends a lead that the Democratic candidate already has, comparable to what occurred in two statewide races in 2018. Or whether, instead, it cuts into a lead that Trump starts with on Election Night—and, if so, whether it is enough of a gain for Trump’s Democratic opponent to overcome Trump’s Election Night lead.

That was all before the coronavirus. Elections held during the pandemic so far have revealed a slew of problems: not enough poll workers, not enough polling stations, and long lines to vote. That’s all happening in relatively small primary elections. Extend that to a nationwide election with the high turnout expected in November and the risks become greater.

Much of that turnout will not be in person. The use of mail-in voting has expanded massively, as states seek to offer voters a way to cast their ballots without having to worry about their health. Although some states use universal mail-in balloting, no national election has ever relied so heavily on it, and most states don’t have experience processing so many mail-in votes.

The explosion of mail-in voting could enormously magnify the blue shift. In the past, there’s been little correlation between mail-in ballots and the post–Election Day Democratic gain. But there is growing evidence, and concern among GOP strategists, that the president’s crusade against mail-in voting is discouraging Republicans from casting their ballots that way. If mailed ballots are disproportionately Democratic, and Republicans disproportionately vote on Election Day, then the blue shift could be huge—especially in states where officials are restricted from counting mailed ballots until Election Day, including Pennsylvania, Wisconsin, and Florida.

Other states might not be presidential tipping points, but could have pivotal House and Senate races. (Two New York Democratic congressional primaries saw winners officially declared only last week, more than a month after Election Day.) Courts and state governments could alter these rules—for example, by mandating different schedules, or requiring that votes postmarked but not received by Election Day be counted.

“We saw delays in the primaries,” Hasen, the election-law expert, told me. “There’s going to be much more volume [in the general election]. There’s a lot that could be done to help. It’s just a question of whether the system will be adequately resourced.”

Even a well-prepared system will see a blue shift, though, and that will make it vulnerable to the kind of attacks Trump used in 2018.

“The public is more likely to perceive, rightly or wrongly, that ballots counted for the first time after Election Day are more susceptible to partisan manipulation than ballots counted on Election Day, with this perception stronger if these overtime ballots tilt more favorably toward one party and diverge from the Election Day count,” Foley and Stewart wrote.

Neal Kelley, the registrar of voters in Orange County, told me it was hard to even describe his experience of trying to canvass voters after the midterms. “It’s just a tremendous amount of pressure,” he said. “You’re getting scrutiny from attorneys, candidates, media, across the board. It’s kind of like captaining a ship in battle.”

Kelley said he doesn’t worry about fraud, noting that in most states there are more checks on mail-in votes than on in-person votes. But he said offering full transparency was helpful in instilling confidence in the tally. TV crews rolled tape all day as his team counted votes, and political operatives were watching too. “I don’t expect people to just blindly trust us,” he said.

Open communication from election officials will be essential to maintaining legitimacy if blue shifts significantly change the initial projections this November. The public has to understand what to expect: not only that the timeline for results will be extended, but also that the final tally might be different from the early returns. But much of the burden for protecting the credibility of the voting will fall on elected officials. Unfortunately, the president’s track record makes it clear that he will cry fraud no matter the result—he did so even in victory in 2016, insisting that “millions of people” voted illegally.

“A lot of this is on responsible members of the Republican Party,” Hasen told me. “If there’s no way Trump’s going to win, even with claims of fraud, I expect Republicans will reject Trump’s claims of fraud. If it’s close, then they’ll get in line behind Trump.”

If that prediction comes true, it might be the only orderly thing to happen this November.

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