The Memo: In defense of the 2020 presidential polls

First impressions matter, and that has been bad news for opinion pollsters this year.

On Election Day and its immediate aftermath — when President Trump looked to have a real chance of winning a second term — pollsters were excoriated for being “wrong,” just as they had been in 2016. 

But as time has gone by and more votes have been counted, a different, more favorable picture has emerged. Overall, pollsters have had a better year in 2020 than they did four years before — though there are important caveats.

There were, to be sure, big misses in some important states. Some polls in key congressional contests, such as in the Maine Senate race, were way off. There was also an overestimation of President-elect Joe Biden’s advantage nationwide. 

The final RealClearPolitics (RCP) national polling average had Biden up by 7.2 percentage points. As of Monday afternoon, Biden’s lead was exactly half as large, at 3.6 points.

A miss on that scale is far from historically unprecedented, however. In 2012, then-President Obama defeated GOP nominee Mitt Romney by 3.9 percentage points, having held only a thin 0.7-point lead in the final RCP national average.

Terry Madonna, the director of Pennsylvania’s Franklin & Marshall College Poll, said that although “there is huge skepticism among the public where polls are concerned,” nevertheless “there is a general consensus among analysts and pollsters themselves that this is largely in line with past years.”

Much of the criticism back in 2016 stemmed from state-level polls that were significantly off in the key battlegrounds of Michigan, Pennsylvania and Wisconsin. Trump won all three, despite polls projecting victory for Democrat Hillary Clinton.

But Biden flipped all three back into the Democratic column this year, and the polls were, with the exception of Wisconsin, much better overall.

Pennsylvania was especially notable. The final RCP average in the state indicated that Biden was leading by 1.2 points. On Monday afternoon, his lead was 1 point. 

In Michigan, the RCP average had Biden leading by 4.2 points and he is leading in actual returns by 2.6 points — a modest difference.

The polling performance was also far stronger than has been credited in other battleground states. 

The final RCP average was within a single percentage point of the real results in Arizona and Nevada and varied by just a fraction over 1 point in Georgia and North Carolina.

Nate Silver of data site FiveThirtyEight wrote last week that he did not “entirely understand” the argument that the polls had been badly wrong. 

“This year was definitely a little weird, given that the vote share margins were often fairly far off from the polls (including in some high-profile examples such as Wisconsin and Florida). But at the same time, a high percentage of states (likely 48 out of 50) were ‘called’ correctly, as was the overall Electoral College and popular vote winner (Biden). And that’s usually how polls are judged: Did they identify the right winner?”

At the same time, Silver alluded to two states where there were significant misses. 

Many polls had predicted an easy win for Biden in Wisconsin. The final RCP average in the Badger State had Biden leading by 6.7 points. 

Reputable organizations, such as Reuters-Ipsos and The New York Times-Siena College, had the Democrat leading by double-digits in the final stretch. An ABC News-Washington Post poll in the final two weeks of the campaign put Biden’s Wisconsin lead at 17 points.

Biden eked out a victory in the state by less than 1 percentage point.

The situation in Florida may have had an even bigger impact on public perception, given the state’s outsize importance. 

The final RCP average had Biden leading by 0.9 points in Florida and, again, several big names in polling gave him a bigger lead than that in their final surveys.

The final Florida polls from Quinnipiac University and Emerson College, for example gave Biden leads of 5 points and 6 points, respectively. With more than 98 percent of the votes counted in the state, Trump’s margin of victory stands at 3.3 points. 

Polls can also get the result right but the margin significantly wrong. There is some evidence that this could be a particular problem in parts of the Midwest, where Trump’s support was markedly understated in polls. 

In addition to Biden’s much closer than projected win in Wisconsin, the polls forecast near dead-heats in Ohio and Iowa. 

Trump was ahead by 1 point in the final RCP average in the former and by two points in the latter. He appears to have carried both states by about 8 points. 

In Iowa, however, the Des Moines Register poll, conducted by Selzer & Company, once again bolstered its reputation as the gold standard, almost nailing Trump’s margin of victory. That poll’s 7-point lead for Trump appeared an outlier at the time, but was vindicated.

Democratic pollster Mark Mellman, who is also a columnist for The Hill, contended that some of the criticism of polling had been “exaggerated.” 

He acknowledged that there had been some underestimate of Trump’s strength — and GOP strength generally — but he noted that some of this could be explained by late-deciding voters moving the president’s way. 

Exit polls bear out this thesis to some degree, and Mellman argued it was a more plausible explanation than the often-heard argument that there are “shy Trump voters” not picked up by polls.

There are clearly issues that pollsters will want to examine in the weeks to come — not least projections for congressional Democrats that proved far too sunny this year.

But they can also build a far stronger case for their own defense than looked likely on election night.

The Memo is a reported column by Niall Stanage, primarily focused on Donald Trump’s presidency.