Why pollsters underestimated the size of the Democrats’ US midterm ‘blue wave’

{"text":[[{"start":9.96,"text":"Pollsters predicted Texas’s Senate race would be one of the closest in this week’s midterms, with James Talarico expected to scrape a narrow lead over his Republican opponent Ken Paxton."}],[{"start":21.88,"text":"The contest in the Lone Star State was one of the most widely watched Senate races in the country — with dozens of polls commissioned in the final weeks of campaigning and hundreds of millions of dollars spent in support of each candidate. Yet on the night, Talarico swept to victory with a commanding 12-point lead."}],[{"start":41.12,"text":"The scale of the win was just one example of how pollsters underestimated Democrats in virtually every competitive Senate race this election cycle — and misjudged the scale of the Republican collapse."}],[{"start":null,"text":"

"}],[{"start":52.08,"text":"Election polling involves “weighting” — adjusting raw responses to match the demographics of the electorate, rather than the demographics of those who responded to a poll. But this is a moving target: determining what the electorate will look like before an election takes place involves modelling and guesswork."}],[{"start":70.52,"text":"“What makes this era of politics so unprecedented is that it is even more difficult to predict the electorate,” said Ashley Koning, director of the Eagleton Center for Public Interest Polling at Rutgers University."}],[{"start":82.84,"text":"“The electorate before an election is always an imaginary population anyway. We don’t necessarily know who’s going to turn out to vote from one election to another.”"}],[{"start":92.72,"text":"Pollsters have underestimated Republican strength in recent presidential cycles. In 2024, 2020 and 2016, polls in swing states consistently miscalculated support for Donald Trump."}],[{"start":null,"text":""}],[{"start":105.5,"text":"Between presidential campaigns, polling companies adjust their methods to correct their mistakes. After underestimating Trump’s support in 2020, for example, many blamed “non-response bias” — essentially that the president’s supporters were less likely than Democrats to pick up the phone for a pollster."}],[{"start":122.2,"text":"But that adjustment may have caused them to underestimate Democrat strength in two straight midterm elections."}],[{"start":128.32,"text":"Pollsters this year began more frequently adjusting their results based on self-reported 2024 votes in an effort to better capture Trump voters. This may have caused them to miss the size of the Democratic swing by assuming the president’s supporters would turn out in higher numbers."}],[{"start":143.4,"text":"For example, South Carolina was widely considered out of range for Democrats but they won by a comfortable 4.4 points. Mississippi was believed to be such a Republican stronghold that it was barely even polled, but Democratic candidate Scott Colom is slightly ahead of his Republican rival in a contest that will be decided by a few hundred votes."}],[{"start":164.22,"text":"The biggest Senate misses were in Georgia and North Carolina, where the Democrats were expected to win but polls underestimated the final margins by more than eight percentage points, according to an FT analysis of data from aggregator FiftyPlusOne and provisional vote tallies from the Associated Press."}],[{"start":null,"text":""}],[{"start":180.7,"text":"Predictions for the House of Representatives were off by similar margins. While pre-election polls improved for the Democrats over the course of the campaign — with their lead increasing from five to nine percentage points over the six months before election day — the final result is expected to be much higher."}],[{"start":196.94,"text":"With vote counts continuing in many states, by Thursday afternoon forecasters and bettors believed Democrats would end up with a roughly 14-percentage-point advantage overall."}],[{"start":206.26,"text":"Despite the errors, Koning said this year’s midterm elections were good for “quality” pollsters with rigorous methodologies; partisan and lower-quality outfits were less successful, with some found to have published entirely fabricated results."}],[{"start":219.76,"text":"“There is such a proliferation of polling and it’s not necessarily up to the same standard,” she said. “Going forward for the polling industry, it’s about the quality of polls that are out there and not just how many polls are bombarding the scene.”"}],[{"start":237.92,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1793933390_2895.mp3"}