To many political observers over the past year, the prospect of President Donald Trump’s reelection looked doubtful at best. Former White House chief strategist Steve Bannon reportedly gave Trump a 30 percent chance of finishing his term. Mike Murphy, the longtime GOP consultant and NeverTrumper, said that Trump will only be president “until early 2019. And JPMorgan Chase CEO Jamie Dimon, like many others, bet more modestly: that Trump would be the first one-term president since George H. W. Bush.

These predictions were understandable. In 2017, it seemed, every day brought another Trump scandal, a bombshell in the Russia story, a brawl in the White House, or a diplomatic crisis, which explains why Trump’s approval rating last year was a historic low for a president’s first year in office. By year’s end, Democrats were anticipating a sweeping victory in this year’s midterm elections, with the opportunity to take back the House and perhaps even the Senate, and an enormous field of candidates was unofficially lining up for the party’s 2020 primary.

But recent data should trouble them. Internal polling by the Democratic group Priorities USA showed the president’s approval rating had climbed to 44 percent in early February, which “mirrors Trump’s improving position in public polls.” Gallup finds a narrow majority of Americans support his handling of the economy, and the new Republican tax law is getting more popular.

“I think people just kind of assume he’s a goner,” FiveThirtyEight statistician Nate Silver told me recently, “but look, he’s now more in a range where presidents have recovered to win reelection. His approval rating is up to 41 or 42 percent in our tracking. That verges on being a normal number that resembles what happened to Reagan or Clinton or Obama in their second years.” (Silver noted over the weekend that Trump dipped to 39 percent in their tracking.) As Jim Messina, who managed President Barack Obama’s reelection campaign in 2012, warned earlier this month, “Donald Trump can absolutely be reelected.”

Former George W. Bush speechwriter Matt Latimer is even more confident of the president’s chances. “Donald Trump is on track to win reelection,” he argued in Politico Magazine over the weekend. “He’s cut taxes. He’s rolled back regulations. He’s put ISIS on its heels. The economy and the stock market are humming along again, despite recent turmoil.”

No analyst I interviewed would speak as confidently. “Only an an amateur would try to predict the results of a presidential election three years from now,” Roger Stone, the longtime GOP consultant and Trump confidant, told me. But the notion that Trump won’t make it to 2020whether because he’s impeached, he resigns, or worse—looks increasingly misguided. So does Democrats’ confidence in taking back the White House. “When you think you’re destined to win,” said Larry Sabato, the director of the University of Virginia’s Center for Politics, “you’re halfway to losing.”

Some analysts have been saying for almost a year that Trump could be reelected. Last May, Columbia University sociology fellow Musa al-Gharbi wrote for The Conversation that “Trump will most likely be reelected” due to the “default effect” in presidential politics: Most incumbents win a second term. He also downplayed Trump’s unpopularity. “Trump won his first term despite record low approval ratings, triumphing over the marginally less unpopular Hillary Clinton,” he wrote. “He will probably be able to repeat this feat if necessary.”

It seemed hard to believe at the time. Then, as now, Trump’s approval rating was in the high 30s, and his presidency was in chaos. He’d just fired former FBI Director James Comey and—the day after al-Gharbi’s piece ran—told NBC News that Comey’s firing was because “this Russia thing with Trump and Russia is a made-up story.” Trump seemed to be in a hurry to get indicted or impeached. But al-Gharbi’s argument is much more believable nine months later.

“If you had the election literally today, I think Trump would be an underdog in the popular vote, but I don’t know about the electoral college,” FiveThirtyEight’s Nate Silver told me.He’s coming from a low point where he had, approval ratings wise, by far the worst first year of any president. But he’s kind of reverting to some mean, in a way, and the mean is how, on average, incumbent presidents get reelected. You know, on average incumbent presidents are having a rough time two years in and their parties suffer anywhere between mild and humongous loses in the midterms, but the baseline case is that incumbent presidents usually win.”

Trump’s most obvious strength heading toward 2020 is the enduring and enthusiastic support of his base. “I would say the most distinctive thing about him other than his obnoxiousness is that his followers aren’t a base,” Sabato said. “They’re a cult. This is a cult. They’ve ceded their independent thinking to this man. This is the most intense cult that I can remember in American politics.” Though their intensity might not be apparent in this fall’s midterm elections, because Trump isn’t on the ballot, they’ll likely show up in massive numbers to support Trump’s reelection in 2020. “If Trump isn’t removed from office and doesn’t lead the country into some form of global catastrophe, he could secure a second term simply by maintaining his current level of support with his political base,” Democratic strategist Doug Sosnik wrote in The Washington Post in October. “Since Trump’s inaugural address, his focus has been on maintaining his support among this loyal base rather than expanding it. As counterintuitive as it may seem, this could be a winning political strategy.”

Sosnik argued that Trump can’t win a two-person race with this strategy, but wrote that the president might have another pathway: “The lack of voters’ faith in both parties increases the probability that there will be a major third-party candidate on the 2020 ballot. It will also lead to other minor-party candidates joining the presidential race. The multi-candidate field will further divide the anti-Trump vote, making it possible for him to get reelected simply by holding on to his current level of support.”

One prominent Democratic strategist told me Trump is most likely to win if he runs as an independent candidate himself. “I think he’s positioned himself from the beginning to run outside the Republican Party, and frankly I think that’s his best option,” said Tad Devine, who served as Bernie Sanders’s senior strategist in the 2016 presidential primary. (Devine thinks Republicans might distance themselves from Trump after he costs them dearly in the midterms.) “Path number two is that the country moves along for three years and continues to create jobs, and there’s no new war that breaks out, and he wins the Republican nomination without contest, and the Democrats have a long and bitter fight,” he said.

Devine isn’t particularly worried about the latter possibility. He calls Trump “the greatest unifier of the Democratic Party,” and said, “I don’t think there will be a problem for Democrats to get behind whoever wins the nominating process.” But not everyone is so sure. “What are the chances you’re going to have that many Democrats fighting among themselves and not have permanent splits, rifts, and divides in the general election?” asked Sabato. “Just look at Hillary and Bernie.” Sabato also noted the third-party factor, telling me, “Inevitably, they’ll take far more votes from a Democratic candidate than Trump.... The more options you have to vote against Trump, the worse it is for the Democratic nominee.

Trump could be hurt, however, by a strong challenge from within the Republican Party. As American University historian Allan Lichtman has argued, “an internal nomination contest” is “the single best predictor of presidential election results.” (Ohio Governor John Kasich reportedly is weighing a Trump challenge in 2020, either as a Republican or independent.) Lichtman, who has correctly called the outcome of every presidential election since 1984 except one (2000), bases his predictions on 13 “keys” including incumbency, scandal, charisma, the economy, foreign policy, social unrest, and third-party candidacies. “If Trump wins reelection,” Lichtman told me, “it will be because he quells a revolt in his own party, establishes more of a conservative agenda than just tax cuts, avoids a big foreign policy disaster, and perhaps even achieves a foreign policy success.” At the moment, he said, Trump is “in a precarious position, but the keys are not so firmly aligned that you can make a clear prediction.”

Lichtman said the only variable that Democrats can control in the 2020 election is whether they nominate a charismatic candidate. He noted that the party’s past three presidents—Barack Obama, Bill Clinton, and Jimmy Carter—were all elected as “young unknowns,” which does not describe the top likely candidates to run against Trump in 2020: Joe Biden, Elizabeth Warren, and Bernie Sanders. The Democratic Party increasingly is looking like a nursing home,” Sabato said.

The conventional wisdom says that a strong economy favors the incumbent party in the White House. That didn’t hold true in 2016, but Sean Trende, a senior elections analyst for RealClearPolitics, thinks the economy will play a key role in how low-information voters evaluate Trump. “The people who really care and are apoplectic or think he’s the greatest guy ever are already voting Democratic or Republican,” he said. “What you’re left with as deciders are these people who don’t follow politics really closely but do know if they got a raise this year of if their cousin got laid off.”

This kind of analysis could help Trump, assuming the economy continues to improve. Silver described precisely how voters could rationalize reelecting the president in such a scenario: “You know what? I was worried about Trump when he was first in office and I don’t like the tweeting, but the fact is things have worked out okay for me personally, the economy seems to be in as good shape as it’s been in many years, and so why not? Why not give him another four years? Nothing blew up.”

Trende told me that “if the election were held tomorrow, Trump would lose,” but warned against overconfidence: “I kind of feel like I’ve seen this movie before.”

Correction: A previous version of this article stated that Allan Lichtman has correctly called the outcome of every presidential election since 1984. He predicted that Al Gore would win in 2000.