The polls are predicting Donald Trump will lose in November — again. Is it worth paying them attention? Pollsters have adjusted their models to try to avoid the mistakes of 2016, but the COVID-19 pandemic leaves little room for certainty.
The problem in 2016 was an overall lack of polling in battleground states — especially polls with more rigid, proven methods. Courtney Kennedy, director of survey research at Pew Research Center, told The Spectator she’s noticed a significant increase in battleground state polling from reputable organizations.
‘This doesn’t mean that the state polling in 2020 is infallible or that the small leads in some states are sure sign of victory,’ she said. ‘But it does suggest that in general, the quality of state polling looks to be somewhat better this cycle than in the last one.’
A common error in swing-state polling from 2016 was an overrepresentation of voters with college degrees, which led to an overestimation of Democratic Party voters. Kennedy said pollsters in this election cycle have emphasized weighing surveys in a way that better reflects the educational background of voters. The key to weighing polls is predicting turnout for specific demographics, which can vary election to election.
‘While this issue is important, it’s not the only reason there were problems in 2016, nor is it the only thing pollsters need to attend to in order for their poll to be as accurate as possible,’ she told The Spectator. ‘A poll also needs to be representative with respect to geography, age, race, ethnicity, urbanicity, sex and potentially more.’
Joe Biden currently holds a steady lead over Trump in national polling — mostly a 4 to 8 percent cushion depending on the poll. The former vice president leads in nearly all swing-state polling too despite Trump having cut the margin towards the end of the summer. North Carolina and Florida are practically tied, Arizona and Pennsylvania are within a few percentage points, and Michigan and Wisconsin are for the most part beyond the margin of error.
Clinton led by as many as 7 percent nationally in mid-October of 2016 before Trump cut the margin to 2 percent in November. So is September polling even relevant? Karlyn Bowman, a senior fellow at the American Enterprise Institute who focuses on public polling data, said there is still plenty of reason to be paying attention to the polls as the election continues to heat up.
‘The closer you get to election day the more polls can tell you — usually they start having greater validity in the summer,’ she told The Spectator, noting that the most recent ABC/Washington Post poll in Arizona, which has Trump leading, started interviewing registered voters rather than likely voters. ‘They use screening questions to find out who will actually vote, and I always watch for the switch which generally happens around this time.’
Kennedy expressed more hesitancy to rely on polling data to predict elections. Respondents, she noted, often change their minds or decide not to vote. She said it’s better to view polls as a flow of information regarding a specific topic rather than a pattern of facts.
‘I do not think it is wise to use polls for predictions at any point,’ she told The Spectator. ‘Polls do, however, give us useful information about how the public is reacting to the pandemic, the economy and the two main candidates. So there can be a lot of value in polling – even in September – but it is for information about issues motivating voters.’
[special_offer]
There’s also, of course, a global pandemic to keep in mind. Fear of spreading disease at polling sites on election day could lead to shifts in voter turnout. An increase in mail-in voting likely means more people will be casting their votes on different days than in past elections. So not only do pollsters have to adjust from their mistakes in 2016, they have to worry about potentially unprecedented voting trends in 2020.
‘The pandemic injects a tremendous amount of uncertainty into pre-election polling and, in some ways, the election itself,’ Kennedy said. ‘I do not have the impression that anyone can predict its influence on polling.’
