How Much Should We Trust the Polls?
Republicans could have a very bad election. After three presidential elections in which polls underestimated Republican performance, we should be less certain about exactly how bad it will be.
Republicans have plenty to worry about heading into November.
Democrats are leading the generic ballot, President Trump’s numbers are weak, voters remain frustrated with the cost of living and the war with Iran has added another political problem at exactly the wrong time.
Republicans could have a very bad election. But how much should we trust the polls telling us how bad it will be?
We have now been through three presidential elections in which the polling industry underestimated Republican performance relative to Democrats.
According to the American Association for Public Opinion Research, polls overstated Democratic margins by an average of 3.1 points in 2016, 4.6 points in 2020 and 2.7 points in 2024.
Those aren’t insignificant misses. A three- or four-point error can turn what looks like a comfortable lead into a close race and can easily put the wrong candidate ahead in polling of a competitive state.
Remember Iowa in 2024.
Three days before the election, the highly regarded Des Moines Register/Selzer poll showed Kamala Harris leading Donald Trump by three points. Trump won Iowa by more than 13.
Some pollsters have done a better job identifying Republican strength.
Trafalgar saw Trump’s support in key states that much of the polling industry missed in 2016, and a RealClearPolitics analysis later found it had the lowest average error across the 2016 and 2020 presidential elections.
Rasmussen’s final national poll in 2024 had Trump ahead by three; he ultimately won the popular vote by about a point and a half. Rich Baris and Big Data Poll are also worth watching because their modeling of the Trump electorate has frequently produced numbers outside the conventional polling consensus.
There are legitimate reasons for these differences. Modern polling requires pollsters to make assumptions about who will vote, how different groups should be weighted and how to reach voters who increasingly don’t answer their phones or respond to surveys.
Trump also brought working-class and lower-propensity voters into the Republican coalition, making an already difficult electorate even harder to model.
None of this means Republicans should dismiss bad polls. When survey after survey shows voters unhappy about affordability or moving away from the party, candidates should pay attention and campaign accordingly. Assuming another polling miss will rescue them would be foolish.
For the rest of us, a healthy dose of skepticism is warranted.
Look at the averages and the trends, but also look at who conducted the poll and that pollster’s record. Pay particular attention to those who have successfully picked up Republican support that others missed.
Republicans may indeed be headed for a rough November. After three presidential elections in which polling underestimated Republican performance relative to Democrats, we should probably be a little less certain about exactly how rough it will be.
The voters still get the final say.