Harry Enten taught me something I didn’t know. Research shows that “from 1988 to 2012, asking people who they believed would win [a presidential election] was more accurate than polls of voter preference of who they wanted to win.” Obviously, that streak ended in 2016.

So, now, the public is giving two very different opinions on these two issues to pollsters. Joe Biden is persistently ahead in both national polls and in key battleground states important for the Electoral College outcome, but people still feel pretty strongly (avg. of polls is 55 percent) that Trump will win reelection.

As Enten notes, this division was evident in 2018:

Gallup asked Americans just before [the 2018 midterms] whether they thought Democrats or Republicans would control the US House after the election. By a 50% to 44% margin, they said Republicans. This came even as Democrats were clear favorites in pretty much every forecast and when that same Gallup poll showed Democrats with an 11-point lead on the generic congressional ballot.

This 2018 poll marked the first time that Americans incorrectly forecasted who would win the House. Americans had previously correctly called every other House election and flip of the chamber in the years Gallup polled (1946, 1948, 2006 and 2010).

Some things we can probably eliminate as possible explanations. Jimmy Carter and Poppy Bush lost reelection, but the people expected that outcome so an assumption that incumbents will win can’t by itself explain why people have a split opinion today. Perhaps people feel like something has changed in the interim, perhaps the existence of right-wing media like Fox News or maybe the impact of the Citizens United Supreme Court case. Maybe they just think Trump is more likely to cheat or that foreign powers intervening on his behalf will put him over the top. It could be that people lack confidence in Joe Biden’s campaign skills, although he just won the nomination almost solely due to his perceived electability. Clearly there is a hangover from 2016 that prevents people from having confidence in the polls.

It’s probably a combination made up of small bits of all these considerations, but the result is that people want Trump gone but don’t believe he actually will be gone. There’s even a whole genre of skepticism that doesn’t believe Trump will voluntarily leave office even if he is defeated.

Yet, people felt similarly in 2018. They doubted the Democrats would win the House even as they gave them an 11-point preference over the Republicans. And it turned out that the House Democrats won in a very convincing fashion.

One thing I think is safe to say is that Trump has a history of doing better at the ballot box than he does in surveys. I guess a lot of people support him without being willing to admit they support him. His supporters are highly skeptical of experts and media, so their cooperation level with pollsters is low, which may also explain some of the disparity. For these reasons, I believe it’s more likely that the polls are skewed against Trump again than that the example of 2018 should give us new confidence in public surveys.

Having said that, if the election were held today, I’m about 99 percent confident that Biden would win, possibly in a shocking landslide that includes states like Texas. But the election is not being held today.