Should We Trust The Polls?

Mark Colley
​
The 2020 campaign is the worst-case scenario for President Donald Trump and his campaign. An unprecedented pandemic, a global recession and a fairly evasive opponent. Could anything go worse?

In the last three weeks, it has. Trump’s performance on the debate stage in Cleveland was largely panned, giving Biden a bump in a race that has been largely stagnant since Bernie Sanders withdrew from the Democratic primary in April. Polling also showed a dip for Trump after his positive COVID-19 test.

The polls have been dismal for the president. According to FiveThirtyEight’s national polling average, Biden leads Trump by 10.4 percent. And the average, which is typically slow in responding to polls dramatically above or below what it thinks is the mean, has seen a number of Biden +14 polls in recent days. (It has also seen many Biden +6 and +7 polls.)

But the polls were wrong in 2016. Pollsters tended to oversample college-educated voters, while Trump’s largest support group — white, non-college-educated voters — were under sampled.

​

There was also a dearth of polling in important swing states, like Michigan, Wisconsin and Pennsylvania. In 2016, polls of Wisconsin were off by eight points. Polls of Pennsylvania were off by five, and polls in Michigan were off by four.

Should we — and can we — trust the polls that show Biden up by 10 points nationally if those same polls made dramatic and fatal mistakes in 2016?

Yes, we can trust them — and we should. Here’s why.

Polls may have been wrong in swing states, but not nationally.

While there were four-, five- and eight-point polling errors in 2016 in some key swing states, the national picture was largely correct. Pre-election polling showed Hillary Clinton leading by three points. She won the popular vote by 2.1 percent.

Pollsters have learned their lesson.

Changes have been made from 2016 to 2020 to adjust for the errors in some of the most vital parts of the country for pollsters — Michigan, Wisconsin and Pennsylvania. Methodologies have been adjusted for the fact that pollsters undersampled white, non-college-educated voters in 2016, and the one example we have of these changes, from the 2018 midterms, shows that they largely worked.

Even if the polls are as wrong as they were in 2016, Biden would still win.

Those four-, five- and eight-point polling errors are pretty scary, and it meant the difference between a Clinton presidency and a Trump presidency in 2016. But in 2020, it doesn’t.

Biden has a large enough lead that even if the polls are as wrong as they were in 2016 — which they likely won’t be, because of changes in pollster methodologies — he would still win Wisconsin, Michigan, Pennsylvania, Arizona and Florida, according to a New York Times analysis. He could also win Georgia and North Carolina.

Even including the polling errors of 2016, these aren’t states that Biden would be winning by the skin on his teeth. He would only have a one-point lead in Wisconsin and Florida, but would have a four-point lead in Michigan, a three-point lead in Pennsylvania and a three-point lead in Arizona.

The “silent majority” might still exist, but its impacts would be baked into the polls.

While this author doesn’t buy the excuse of the “silent majority” supporting Trump, let’s assume that there is a group of voters too shy to show support for the president in polls. (Even though polls are entirely private and anonymous, and the “silent majority” is pretty much just polling error, anyways.)

This was the same argument used in 2016, and as enumerated, the numbers are far better this year than they were for Clinton, even when you account for the polling error in 2016. It’s hard to envision a scenario where the silent majority has grown since 2016.

Biden is a very different candidate than Clinton.

Besides the quantitative argument to be made for Biden in this election, there is a qualitative argument as well. Joe Biden is a vastly different candidate from Hillary Clinton, in that he is nowhere near as hated and hard to pin down with the same scandals that drove Clinton mad in 2016.

Sure, there’s the potentially-sketchy dealings in Ukraine and numerous legislative mistakes in Biden’s past, but Trump has failed to be as effective in creating a media circus around these criticisms as he was in 2016. Perhaps it’s because the media is simply not covering these issues as it did in 2016 — or perhaps it is because of better favorable numbers for Biden.

The Bottom Line

Joe Biden is not yet out of the woods. Even in three weeks, so much can change — 2020 has shown us that much. There are still moments of vulnerability ahead for the former Vice President’s campaign, and the wheels could still come off.

Donald Trump still has about a 1-7 chance of winning the election. It’s not impossible to envision a world where he receives a bump from the Supreme Court nomination, pushes through a heroic last-minute piece of vaccine news, or even finds some effective messaging on Biden that finally does what he did to Clinton in 2016.

But as the days tick down, it is increasingly unlikely that this will happen. For many reasons, we can and should trust the polls that show him with the large lead that he has.

To simply say, “The polls were wrong in 2016, so they’re wrong now,” is cavalier and ignorant to the more nuanced differences between the two elections. Things are different now, and the polls show it.

​

---

Read More 990WBOB

• Pats PostGame Notebook: Bucked by the Broncos
• Taste Test: Wendy's Pretzel Bacon Pub Cheeseburger
• Giberti: Stop Lying to Us​
• Will A Third Party Ever Succeed in America?