WASHINGTON – Predictions based on survey results that Hillary Clinton would win the 2016 presidential election spurred widespread distrust in the polling industry. Pollsters adapted to learn from their mistakes, but several state-level prediction faults in this year’s presidential polls ignited criticism and led to questions about the future of the industry.
It won’t be clear how the national polls held up compared with the final totals until all the votes are counted. On a state level, it appears many polls overestimated former Vice President Joe Biden’s support in Florida, Ohio and Iowa. These states were predicted to produce narrow Biden wins, yet swung to Trump by margins of several percentage points.
“When polls turn out to be as wrong as they are, then people lose trust in the actual polling system itself. And as a part of that, they start to lose trust in the system,” Lawrence Stuelpnagel, political science and journalism professor at Northwestern University, said.
Courtney Kennedy, director of survey research for the Pew Research Center, said in a report that a “robust public polling industry is a marker of a free society.” In a democratic nation, this is an indicator of how well organizations can gather and distribute information about the public’s opinion, she said.
Several political figures and journalists have said on Twitter that this year’s pre-election polling mistakes mark the end of the polling industry.
Trump doing better than was predicted by the vast majority already… The big loser tonight are the pollsters and most of the media who no one will ever believe anymore.
— Meghan McCain (@MeghanMcCain) November 4, 2020
“The political polling profession is done,” leading Republican pollster Frank Luntz tells Axios. “It is devastating for my industry.”
— Ben Tracy (@benstracy) November 4, 2020
Polling drives coverage. Polling drives fundraising. Polling helped drive the Democratic primary by determining who made the debate stages. Polling drives expectations, and thus faith in the results.
The polling was way off in the primaries and way off in the general.
— Edward-Isaac Dovere (@IsaacDovere) November 4, 2020
Some have attributed these faults to pollsters’ monolithic approach to demographic groups, specifically as seen with Latino voters in Florida.
“Pollsters need to ask themselves: ‘Am I missing those people who are in the crowd …?’” Stuelpnagel said.
A common defense of pre-election polls is that they define the popular vote, not the Electoral College votes. Others have said these polls analyze the public’s attitudes, but not necessarily whether those surveyed will go out and cast a vote based on those attitudes.
Another defense from pollsters is that the data samples are snapshots in time. Pre-election poll results are heavily dependent on timing and have the potential to change before the election.
But Stuelpnagel said it’s as important to ask: “Where are they pointing their lens?”
There is no sole reason for the incongruencies in polls from the past two election cycles. Mail-in and early in-person voting may have contributed to the perception of an early Republican lead when Election Day votes were first counted, inflaming the notion that the polls were wrong.
“While it’s clear that there were some problems with polling this election cycle, it’s too soon to say anything definitive about what those problems were, what caused them, and whether the problems are bigger in scale than in other recent presidential elections,” the Pew Research Center’s Kennedy said in an email.
In the 2016 election state polls for Michigan, Pennsylvania and Wisconsin missed a last-minute swing towards Trump among undecided voters and overestimated the number of college-educated voters who preferred Clinton.