For the third straight election cycle, public opinion polling once again underestimated support for President-elect Donald J. Trump, resulting in a significant polling error. The left-wing Guardian newspaper, based in the United Kingdom, saw their U.S. national polling average err by three points with 99 percent of the vote in the presidential election counted. This error is fairly consistent with the polling misrepresentation of Trump in the past two presidential election cycles.
The Guardian‘s polling average appears to have been thrown off by the inclusion of several polls viewed as outliers by other websites that aggregate public polling. According to their aggregate, the polls predicted a slim lead for Kamala Harris, with 48 percent of the vote to Trump’s 47 percent. However, as the election results stand now, Trump has won a majority of the popular vote at 50 percent, with Harris coming in at 48 percent.
Regarding key swing states, the polling inaccuracies varied. In states such as Georgia, polls aligned closely with results, whereas in places like Arizona, a discrepancy of four percentage points was noted. Interestingly, when looking at individual pollsters regarding the national popular vote, Richard Baris and Big Data Poll were the most accurate pollster of the cycle. They saw only a 0.1 percent error from the final popular vote results. The Wall Street Journal‘s error was only 0.4 percent, while AtlasIntel—the most accurate pollster of the 2020 cycle—was only off by 0.5 percent. The New York Times/Siena poll’s miss was just 0.6 percent, and Rassmussen Reports was off by just. 0.8 percent.
Conversely, pollsters like Marist, Morning Consult, and Quinnipiac saw significant polling errors, and their inclusion in some aggregates likely resulted in the third straight miss on support for President-elect Trump.