In October 2016, we were told Hillary Clinton would beat Donald Trump, whom the political establishment largely considered to be a laughingstock. The New York Times gave Clinton an 84% chance of winning as election night kicked off. And yet, in a polling failure of epic proportions, Trump won by 77 electoral votes. Of course, four years later, Joe Biden, who was leading the 2020 polls by an impressive margin, would go on to deliver the win that Clinton never could. But flash forward to now, and the circumstances for the president are very different, with most polls putting Biden well behind Trump, particularly in the swing states. Which raises the question: Has Biden indeed lost the American public, or are the polls trying—and failing—to measure a shifting electorate?

Many Democrats in Washington fear it’s the former, according to Politico, which reported this week on a “pervasive sense of fear” that “has settled in at the highest levels” of the party. Meanwhile, there are those in Bidenworld, including Democratic strategist Simon Rosenberg, who have sought to allay existential concerns about Biden’s reelection chances, shooting down the notion that the campaign is running on “hopium.” Needless to say, nobody has a crystal ball. But it’s worth pointing out some of the caveats in all of this hand-wringing and hand-waving.

First, it goes without saying that the polling industry does not have nearly the same kind of encyclopedic knowledge or access to people’s political persuasions and habits that tech companies do. For example, it still relies on outmoded forms of outreach, including landline calls, cell phone calls, and text messages—all of which can lead to biases and sampling errors.

And then there are the caveats that are completely unique to this cycle. One of them, much like the “hidden Trump voter” of 2016, is what I’ll call the hidden non-Trump voter of 2024. That is to say, Nikki Haley’s zombie campaign has still received more than 10% of the vote in most of the Republican primaries since its candidate suspended her bid. Some two months after she dropped out, Haley won one in five Indiana Republicans, while Trump’s performance in multiple primaries has been weaker than what was predicted in polls. “In New Hampshire, surveys ahead of the primary showed the former president beating his last remaining rival, Nikki Haley, by nearly 18 percentage points; he won by 11,” the Financial Times reported. “In South Carolina, he was polling 28 points ahead of Haley, but he won by 20. In Michigan, the polls said 57, but he won by a ‘mere’ 42.” All of this makes me wonder whether the pollsters might be overestimating MAGA members’ commitment to actually showing up for their guy.

Another point worth noting, as Times political analyst Nate Cohn recently explained, is that Biden has repeatedly come out ahead in polls of 2020 voters, even if he is falling behind the former president when it comes to registered voters overall. “The Biden hopium argument I mostly do buy is that Biden seems to be doing better among likely voters,” tweeted Nate Silver, “and the likely voter/registered voter gap could be larger than usual in an election people largely feel apathetic about.” If this is true, Trump needs to convince people who don’t traditionally vote to change their behavior and vote for him. That means the former president needs to make the case that he will radically improve voters’ lives—or that Biden will radically do the opposite—despite the fact that voters have already witnessed both a Biden and a Trump administration. It’s quite a heavy lift. But it’s still one that Trump might be able to pull off, considering his aptitude for making large numbers of people believe in bullshit.

While all these caveats may cut against Trump’s ostensible polling lead, that doesn’t mean Bidenworld should sleep easy. Why? For one part, Trump’s electorate appears to be growing: Recent polling indicates that 17% of Black voters would throw their support behind Trump—roughly twice the percentage that did in 2016. This would be a seismic racial realignment.

Last week Trump held a rally in the South Bronx, where he got a fairly big audience (though apparently not quite as big as he claimed). Still, it suggested an opening for Trump among voters who don’t traditionally go for Republicans. Why would members of a predominantly Latino community back the guy whose immigration plan involves mass deportation and potentially detention camps? Hard to say for sure, but dynamics like these pose a major challenge for the Biden campaign.

The other thing that might make Bidenworld shudder is that the former president seems to have a hold on Americans who aren’t tapped into the news cycle to begin with. “Trump has a major lead among those who don’t follow political news,” reports NBC News. “Fifty-three percent back him, and 27% back Biden.” Again, it’s hard to know precisely why this is, but if I were to hazard a guess, I’d pin the blame largely on Big Tech, which has decimated local media and ushered in a political “news” environment chock-full of mis- and disinformation, where someone like Trump can thrive.

We are currently more than 150 days away from the 2024 election. None of us—not even the Nates—know the full extent of what can happen in the homestretch. Bidenworld could still change hearts and minds on both inflation and the war in the Middle East. Trump’s support could falter if he’s convicted in his Manhattan hush money case. Robert F. Kennedy Jr. could, for whatever reason, suddenly drop out. There is always an element of the unknown in every election. But, as past elections have shown, polls should not stand in for wisdom. They are more of an ambiguous art than a perfect science.