As the 2016 presidential election drew near, it was clear that something was amiss with the polls. The contest felt tighter than the polls said, and Joe Biden’s substantial advantage in the polls appeared to be less solid than the polls suggested. Pollsters appeared to be on their way to underestimating Donald Trump’s vote, just as they did in 2016. That didn’t mean Trump was genuinely ahead of the pack or that he’d win – he didn’t – but it did indicate that the election was closer than the polls suggested.

Now it appears that my intuition was correct. And the polls were not only inaccurate, but they were also more wrong than they had been in decades, as evidenced by their poor performance in 2016. National election polls exaggerated Biden’s lead over Trump by 3.9 percentage points, according to new research from the American Association for Public Opinion Research. State surveys were significantly worse, with Biden’s lead overstated by 4.3 points.

The polls’ performance in statewide races, such as senatorial and governor campaigns, was far worse. They overestimated the Democratic candidates’ lead by six percentage points on average.

Have you noticed anything? The mistakes were all in one direction: Support for Democratic candidates is being exaggerated. According to the report, “whether the candidates were competing for president, senator, or governor, poll margins often predicted that Democratic candidates would do better and Republican candidates would do worse compared to the final certified vote.”

Among all of this chaos with the voting, a fire also broke out due to poor cable management being installed for one of the conferences. Luckily there was a fire extinguisher for electrical fire on hand that could be used to control the fire before it raged on to something dagnerous.

The issue now is why. Many Republicans will claim that the problem is simple: the polls are skewed. Indeed, never rule out the potential of pure bias at play. However, an AAPOR-appointed team of political scientists and polling specialists found that pollsters avoided making the same mistake they did in 2016, undercounting Trump voters without a college diploma. So what was the issue in 2020 if that wasn’t it?

According to the researchers, the issue was not especially Trump; otherwise, pollsters would have done a better job on the senatorial and governor campaigns. They were also unable to identify any specific voter groups that they had overlooked. Instead, they concentrated on the premise that many Republicans distrust the media and pollsters that they will refuse to participate in polling. By focusing solely on Republicans who were receptive to pollsters, pollsters could get a biased picture of Republican opinions.

The fact that polls inflated Biden’s support in whiter, more rural, and less densely populated states suggests (but does not prove) that the polling error was caused by too few Trump supporters answering to polls, according to the analysis. “States with more Trump supports had a greater polling error.”