Without question, polling throughout the era of Donald Trump the Politician has been atrociously bad, and now one of the industry’s top experts has come forward to reveal just how bad – as in, slanted against Trump – it’s all really been. In distancing himself from the political establishment, he’s warning of a dangerous collapse in American polling that he says has distorted election expectations throughout the Trump era.
“For the third consecutive election cycle, [polling] has reflected not the electorate as it is, but as the political class wishes it to be,” Jason Corley of Quantus Insights wrote. His new report outlines what he terms a “fingerprint of bias,” indicating that since 2016, polls have consistently overestimated Democratic performance by an average of 4.2 points in key battleground states. “These are not isolated misses. They are consistent, directional, and embedded in the structure of how modern polls are conducted,” he said. In other words, they were ‘polls’ that were not reflective of true voter sentiment, but rather to drive a certain narrative and influence outcomes in favor of Democrats.
Corley highlights a clear pattern: Trump’s wins in Wisconsin, Michigan, and Pennsylvania in 2016; Joe Biden’s narrower-than-expected margins in 2020; and Kamala Harris’s defeats in 2024, despite leading in major polls. In Iowa, for example, the gold-standard Des Moines Register poll had Harris ahead by three points, but Trump ultimately won the state by thirteen.
“This isn’t random error. It’s a fingerprint of bias,” Corley said. And the issue isn’t last-minute voter swings. According to Quantus Insights, it’s structural. “Response rates have collapsed—from 20% in 2000 to just 5% in 2024. The Americans still answering surveys are more urban, more educated, and more institutionally trusting than the electorate as a whole.” Meanwhile, rural, working-class, and conservative voters either can’t be reached or refuse to participate. “The result: pollsters continue to model an America that does not exist.”
Quantus argues the damage extends beyond bad numbers. It is “strategic, psychological, and institutional.” Flawed polling signals false narratives, reshaping campaigns, media coverage, and voter behavior. As Corley notes, “When polls signal a clear lead, voter behavior shifts. Supporters of the presumed winner may stay home. Opponents may lose hope.”
Public confidence in polling has plummeted, with Gallup reporting a drop from 38% in 2000 to just 22% today. Corley argues that a major reason polling failures continue is the influence of aggregators and media figures. “Forecasts of 70% or 80% ‘win probabilities’ are built on fragile foundations—polls that are themselves skewed, but dressed in graphics and given the sheen of certainty.” He then pointed to the now-defunct FiveThirtyEight, whose pollster grading system often “elevated firms with poor predictive records while dismissing newer entrants that showed better performance.”
At Quantus, Corley’s team used a different approach. Rather than relying heavily on polling, they incorporated behavioral models, economic sentiment, turnout simulations, and volatility measures, treating polls as just one of several inputs rather than the dominant factor. Their final 2024 forecast gave Trump a 65% chance of winning 312 electoral votes, projecting a national margin of 48.8% to 48.4%. Trump ultimately won exactly 312 electoral votes, carrying the national popular vote 49.8% to 48.3%.