Another systematic polling error — this time larger than 2016 — didn't quite cost Joe Biden the presidency, but it seems pretty clear that had Republicans done about one point better across the board we would be looking at Republican control of both the House and Senate to go along with a narrowly re-elected President Trump, who would take office despite losing the popular vote by as many as 5 or 6 million votes. As it is, Democrats will eke out control of the House and will fall just short in the Senate unless they can sweep January two runoffs in Georgia. Depending on the outcomes there, Democrats will represent between 20 and 41 million more people than their Republican counterparts.

The worsening disjuncture between the public's desires and the way our institutions translate them into control of American political institutions is an unsettling problem for our democracy and one that continues to exclusively benefit Republicans. But this year's polling miss should also force Democrats to ask some difficult and uncomfortable questions about the last four years.

Chief among them is this: Was Trump more popular than public opinion polling suggested throughout his presidency? Is there in fact a reservoir of Americans that pollsters just can't reach who supported Trump the whole time, mostly sat out the 2018 midterm elections (when polls were more accurate than in 2016 or 2020), and then turned out in droves for him in 2020? And if so, what does it mean?

Votes are still being tabulated, and while we should be cautious and avoid jumping to conclusions before the results are finalized, a clear enough picture is emerging. Biden currently leads in the popular tally by 3 percentage points and 4.4 million votes. Nate Cohn of The New York Times predicts that Biden will end up with a win in the range of 5 points. That's about 3.4 points less than the final Five Thirty Eight polling averages — a big miss but not catastrophic enough to end polling as a trusted industry. But the misfire was worse in some of the key battleground states. Biden led by 8.4 points in Wisconsin, where he will ultimately triumph by less than a point. In Pennsylvania he led by 4.7 and will likely win by about a point. And the averages were off by more than 5 points in Michigan and Florida and over 7 in Iowa and Ohio.

This isn't random chance. The voters who unexpectedly turned out for President Trump on election day existed for the entirety of his term, even if survey researchers never got ahold of them. One explanation that analysts are converging on is the idea that, as Sean Trende argues, what links these voters together is “low social trust” and that this variable can divide otherwise demographically similar groups like non-college educated whites. This kind of voter isn't lying to pollsters as much as they are hanging up on them. And because those voters are already extremely difficult to find and talk to, when pollsters weight their samples based on the ones they do reach, they will still be wrong.

There is almost no other credible explanation for what happened here, because it seems extraordinarily unlikely that these voters disapproved of the president all along, told pollsters as much, but then at the very last moment changed their minds in a way that no reputable organization could pick up. They were there all along, and their absence from public opinion research -- not just Trump's approval but every issue survey researchers have been asking about over the past four years (or possibly longer!) -- probably skewed our overall understanding of the politics of this era.

What number should we hang on the underestimation of Trump's popularity? 4 points? 5? Does it matter? Nearly every piece of analysis I produced over this time period was premised on Trump's deep unpopularity, as measured by thousands of data points over five years since he descended the escalator at Trump Tower and announced his candidacy.

“Deeply unpopular” is a phrase I employed to describe the president more frequently than he uses the word “strongly” as an adverb. But if Trump basically always had a cushion of five points more than the averages suggested, was he ever really “deeply” unpopular? Or was he just particularly polarizing, with perhaps bare majority support at his highest points and dipping into the low 40s (rather than the mid-30s as polls would have it) during especially chaotic and scandal-ridden periods of his term?

Go ahead and scroll through the Five Thirty Eight approval averages for Trump's presidency. Add 2.5 points to his approval and subtract the same from his disapproval. Among other things, what emerges would suggest he was quite close to 50-50 on Election Day itself. Had we known this, it obviously would have made some of the results less shocking than they were.

It also might have changed the Democrats' strategy. Biden made this election a referendum on Trump, and his gamble paid off, although only just so. But the party's congressional candidates were counting on Trump's unpopularity to produce another blue wave like the one they rode in 2018, and it just didn't happen.

That failure suggests they should have been running on more of an affirmative agenda, to tell voters all the cool things that might happen if they gave Democrats control of Washington, and less on a defensive posture of how Trump and the Supreme Court would gut pre-existing conditions or overturn Roe v. Wade. For example, I challenge you to figure out exactly what it was doomed Democratic Senate candidate Sara Gideon was proposing to do with the power she was asking voters to give her in ads like these. For the most part, she let her opponent, Susan Collins, define the policy space for her.

Again, would this have been the strategy they chose if they had known President Trump was a bit more popular than they believed? I don't know, but I have a hard time seeing how she could have done worse than losing by 9 points to Collins in a state Biden carried by roughly that same margin. Maybe those ticket-splitters didn't care for Trump himself but didn't hate some of the things he did in office.

That's because it couldn't have just been Trump's approval that was off. If researchers couldn't get these voters to talk to them about the horse race, they were probably also hanging up when Pew was calling to ask them about their attitudes on immigration, trade, and policing. Pew's data shows that majorities of Americans supported increased deportations of people who are in the United States illegally (while also, confusingly, showing supermajority support for finding ways for them to stay legally) throughout the Trump era. If you apply the 5-point Trump shift to these numbers, it is deportations that have supermajority support.

I don't particularly care for the bleak place where this is all leading — a world where the president's antics and cruelty had much more support than I ever imagined. But part of figuring out how to win is looking directly at the facts and not blinking. Real talk: he didn't just coast on Barack Obama's economy. There were probably parts of President Trump's policy agenda beyond the country's underlying economic performance which had broad support and which had meaningful political contrast with Democratic priorities.

Chief among them was likely his combative trade policies. Not to pick on Pew, but a 2018 report showed just 36 percent of Americans believed that trade with other countries creates jobs. The real number might be closer to 30 percent. Regardless of whether in practice Trump's disruptive trade maneuvering led to suffering and losses for people they were designed to help, it may be that some subset of Americans felt he was standing up for their interests and gave him credit for it. Connecting the dots here, I'm guessing it is Americans with particularly low social trust who respond well to pugilistic appeals to nationalist manufacturing policies.

To be clear: less unpopular than we thought does not mean ‘popular.' A president overseeing a booming peacetime economy should not have struggled to break the surface of 50 percent approval throughout his presidency. And President Trump did, after all, lose the election about as decisively as anyone has in the post-Reagan era of partisan polarization. The fact that more people might have lapped up his divisive rhetoric than we thought does not excuse it, nor should it invite Democratic leaders to behave similarly. It doesn't mean the progressive agenda — many parts of which have high levels of public support even if you shave off a few points here and there — should be treated like a syringe full of coronavirus by national Democrats.

The results do, however, call for reassessment. Democrats have a real advantage here because of what the GOP is doing. Rather than processing their loss and trying to figure out how they can better appeal to Americans who just rejected them for the 7th time in the last 8 presidential elections, Republicans are disappearing down a dark rabbit hole of conspiracies and denial. If Democrats can figure out how to reverse their losses with the voters who evaded pollsters without abandoning their core principles, they might be able to finally overcome the systematic obstacles our antiquated electoral system puts in their way and score decisive congressional majorities for the first time in a decade.