CNN  — 

President Joe Biden won the 2020 election as polls suggested he would. His victory, however, ended up being tighter nationally and in a number of swing states than expected. The question of why that happened has dogged pollsters since the election.

On Tuesday, a number of Democratic pollsters (ALG Research, Garin-Hart-Yang Research Group, GBAO Strategies, Global Strategy Group and Normington Petts) put out a statement as to why they think their internal polls were off. It jibes with a lot of the analysis I and others have said publicly.

Perhaps the biggest takeaway is that the people who aren’t answering polls are systematically different than those who do answer the polls. This is likely the case among White voters without a college degree. Additionally, voters who were less likely to turn out but still did seemed to favor former President Donald Trump at higher levels than expected.

Not being able to accurately gauge the preferences of non-college White voters is a huge problem for pollsters. This group has become increasingly Republican the last few cycles and favored Trump by around 30 points this past election cycle. They make up a larger share of the electorate than two key groups individually: voters of color and Whites with a college degree.

Pre-election polls suggested that Biden would make major gains with this group. It didn’t happen. (Still, polls did a fairly decent job of gaming out what would happen in the rest of the electorate, such as the shift in Hispanics toward Trump.)

My own initial analysis after the election showed that the polling errors tended to be greatest in states where the non-college White share of the population was largest. It was not the case, however, that in most of these places that pollsters didn’t have enough White voters without a college degree in their polls.

Instead, it was likely the case that the people who weren’t answering polls in places where a lot of non-college White voters resided had different political beliefs than those who did answer the polls.

The reasons why the people who picked up the phones differed is up for discussion. The Democratic pollsters suggested two potential reasons that seem plausible.

The first is that voters with low social trust are less likely to take polls. We know from previous research that those voters are less likely to have college degrees and more likely to be fans of Trump.

If you follow this line of thinking, the White voters without a college degree who do participate in polls are more likely to back Democratic candidates than this demographic as a whole. Allowing this subsection to stand in for all White voters without a college degree simply won’t make up for the problem of the polls not capturing those who don’t participate.

The second explanation is that the coronavirus pandemic exacerbated the problem of Biden voters being more likely to answer polls. There had been reports from pollsters that response rates rose because people were stuck inside because of Covid-19.

The people who were stuck inside, however, were more likely to be Democrats. According to a mid-March Axios/Ipsos poll, Democrats were 34 points more likely to say that they were staying home as much as possible because of the coronavirus outbreak. Among those who worked full or part time, Democrats were 20 points more likely to say they had worked at home or remote than their usual workplace.

These general patterns hold, even when controlling for education levels and race.

The last big takeaway was that the people who did turn out and weren’t expected to were far more likely to be Republican than these pollsters thought.

While accounting for every individual pollster’s individual turnout expectations is difficult, the Democratic pollsters’ belief does fit overarching narratives.

A previous study from Equis Research, a Democratic firm, found that Trump made some of his biggest gains from Hispanic voters who usually don’t vote.

The network exit poll also suggests (as I noted previously) that low propensity voters may have actually been more helpful to Trump than is commonly assumed.

Biden won by 1 point among voters 25 years and older, according to the exit polls. I zone in on this group because at least by age, every voter in this group was eligible to vote in 2016.

When we look at just those voters 25 years and older that say they didn’t vote in 2016, Trump won by 3 points.

If pollsters thought that the higher turnout would benefit Biden, they were likely solely mistaken.

The question obviously now is where to go from here. Obviously, pollsters will continue to spend a lot of time looking over the data. Two elections in a row where Trump was underestimated is not a good thing for the pollster profession.

Still, keep in mind that in most races and on most issues, voters aren’t split nearly down the middle. The national polls overestimated Biden’s margin by less than 4 points. If someone or some issue is as popular as Biden is today (about 15 points more of the public approves than disapproves), you can be pretty sure that is real and not an artifact of a polling error.