The 2024 US Presidential election delivered a surprising result, with Donald Trump securing a decisive victory that defied predictions of a close contest against Vice President Kamala Harris. While pundits and pollsters anticipated a nail-biting race, Trump’s triumph was fueled by a surge of support across various demographics and regions, a trend that dramatically contrasted with the prevailing sentiment reflected in opinion polls.
Experts are grappling to understand how the polls missed the mark so significantly. Michael Bailey, a political science professor at Georgetown University, highlights the polls’ inability to capture Trump’s widespread appeal: “They did fine in battlegrounds, but…they failed to provide the essential information that Trump was surging across the board.” This isn’t the first time pollsters have underestimated Trump’s popularity. In fact, according to The New York Times, Trump secured a higher vote count than in 2020 in over 90% of US counties.
The pollsters’ performance has come under intense scrutiny this year, following their failure to anticipate Trump’s 2016 win and their overestimation of President Joe Biden’s margin of victory in 2020. Pedro Azevedo, Head of US polling at AtlasIntel, explains, “Trump was underestimated by about two points this time around in key states.” This underestimation manifested in several key battleground states. In Pennsylvania, RealClearPolitics’ polling average placed Trump ahead by just 0.4 percentage points, yet by Wednesday, he held a two-point lead. Similarly, Trump outperformed predictions in North Carolina, winning by three points despite an expected margin of 1.2 points. In Wisconsin, while polls predicted a narrow 0.4-point lead for Harris, Trump ultimately led by 0.9 points. Overall, Trump’s performance in swing states exceeded poll predictions by an average of three points. He is also on track to win the national popular vote, surpassing his 2016 performance.
The root of the problem, which has persisted since Trump’s entry into the US political landscape a decade ago, lies in a segment of Trump’s electorate who consistently refuse to participate in opinion polls. Pollsters have struggled to accurately assess the impact of these non-respondents. A recent poll conducted by The New York Times and Siena College revealed a significant discrepancy, with data analyst Nate Cohn noting that “white Democrats were 16 per cent likelier to respond than white Republicans.” This gap only widened throughout the 2024 campaign. Despite attempts to adjust for these flaws with statistical methods, pollsters have struggled to bridge the gap.
Azevedo emphasizes the pollsters’ underestimation of Trump’s growth among Hispanic voters, a trend evident in his unexpected victories in Nevada and Florida. Additionally, Trump outperformed expectations among white voters, particularly in rural areas, contrary to predictions that Harris would “improve her margins” in this demographic. Iowa serves as a stark example of these discrepancies. Just three days before Election Day, a poll projected a narrow three-point victory for Harris in the typically Republican state. However, as Azevedo pointed out, Trump won Iowa decisively by 13 points. J Ann Selzer, the author of the inaccurate Iowa poll, attributes the discrepancy to the influence of late-deciding voters and the reluctance of some voters who had already voted to disclose their choices.
The 2024 US Presidential election once again underscores the limitations of opinion polls in accurately capturing the complexities of voter sentiment and predicting election outcomes. While pollsters strive to refine their methods and address inherent biases, the persistent challenge of gauging the impact of non-respondents, particularly among certain segments of the electorate, remains a critical factor in understanding the accuracy and reliability of pre-election polls.
