Polling problems: Fool us once …

Election night in San Francisco.Jim Wilson/The New York Times
It’s another record day for virus cases. Obama’s memoir gets a glowing review. And we look at why the polls were wrong.
The polls were wrong again, and much of America wants to know why.
Dozens of pre-election polls suggested that Joe Biden would beat President Trump by a wide margin, but the race instead came down to one or two percentage points in a handful of states. Polls also indicated that Democrats would do much better than they did in congressional races.
So what happened? Here are six key points:
1. In the last few years, Republican voters seem to have become less willing to respond to polls. Maybe that shouldn’t be surprising, given Trump’s attacks on the media, science and other institutions.
2. This phenomenon isn’t simply about working-class whites. Pollsters were careful to include more of these voters in their samples than four years ago, when the polls also missed, but it didn’t solve the problem. One likely reason: Even within demographic groups — say, independent, older, middle-income white women — people who responded to polls this year leaned more Democratic than people who did not.
3. It’s also not just about Trump. Polls missed in several Senate races even more than in the presidential race, which means they did an especially poor job of finding people who voted for Biden at the top and a Republican lower down the ballot.
4. Most of the easy solutions are probably not real solutions. Since Election Day, some campaign operatives have claimed their private polls were more accurate than the public polls. That seems more false than true. Biden, Trump and both parties campaigned as if their own polls matched the public polls, focusing on some states that were not really competitive and abandoning others that were close.
5. Polls have still been more accurate over the last four years than they were for most of the 20th century. As pollsters get more information about this year’s election and what went wrong, they will try to fix the problems, much as they did in the past. A new challenge: In the smartphone age, poll response rates are far lower than they used to be.
6. We journalists can do a better job of conveying the uncertainty in polls. Polls will never be perfect. Capturing the opinions of a large, diverse country is too difficult. And in today’s closely divided U.S., small polling errors can make underdogs look like favorites and vice versa. All of us — journalists, campaign strategists and the many Americans who have become obsessed with politics — shouldn’t forget this. We just got another reminder.

By The New York Times
For more: I’ve written a longer Times story about this year’s polling errors and the history of the industry.
And my colleague Nate Cohn, who knows more about this subject than almost anybody, points out that a significant chunk of the error involved Hispanic voters. Nate has also discussed polling on episodes of “The Daily” and “The Argument” podcasts.
Elsewhere: Sarah Isgur of The Dispatch says the problem isn’t about Trump voters who lie about their preference. Charles Franklin of Marquette University suggests the pandemic may have affected turnout in surprising ways. Kristen Soltis Anderson, a Republican pollster, notes that polls in many states will still be “incredibly close” to the final result.