Silver Bulletin’s 2024 presidential polling model estimated that some firms’ surveys shifted Kamala Harris’ margin by more than two percentage points against the polling average, a statistical adjustment rather than a finding of political intent.
In a Sept. 27, 2024, analysis, co-authors Nate Silver and Eli McKown-Dawson presented estimates for pollsters with at least five Harris-Trump surveys in the model. Rasmussen Reports’ estimate was −2.6 percentage points for Harris’ margin. Trafalgar Group was at −2.7 points, Spry Strategies at −2.3 and InsiderAdvantage at −0.8. Positive estimates meant Harris’ results at a firm tended to sit above the polling baseline; negative ones put them below it.
Other estimates were +1.9 points for Ipsos, +1.4 for Public Policy Polling and +1.2 for Morning Consult. The New York Times/Siena College was at −0.06 points. Silver used a Rasmussen poll showing a one-point Republican lead to illustrate how the adjustment could make the result look more like a two-point Democratic advantage.
The model used other surveys to establish a baseline for each state or national contest, then recalculated that baseline and the pollster adjustments. Estimates for firms with fewer polls were pulled closer to the average, while more polling data made estimates more precise. The authors said nonpartisan polls were drawn toward zero.
Sponsorship also informed an initial adjustment for partisan-sponsored polls. Silver and McKown-Dawson said those polls typically overstated the sponsor’s candidate margin by about 3 percentage points. The estimate described a statistical pattern, not evidence that a pollster intended to favor a party.
Silver Bulletin later explained that a firm’s difference from other pollsters and the gap between polling and the election result measure separate things. Its post-election review calculated that polls underestimated Republicans by an average of 2.9 points in 2024. The review put Democratic polling bias at 3 points in 2016 and 4.7 points in 2020, with less Democratic bias in the 2022 midterms and slight Republican bias in 2018. The Atlantic’s 2025 review also said polls underestimated Donald Trump by about 3 points and Republican-aligned pollsters performed better. Federal Election Commission results show Trump received 312 electoral votes and Harris received 226.
