Harry Truman’s re-election campaign in 1948 seemed like a losing effort. His approval rating hovered in the mid-30s, and many of his fellow Democrats were jumping ship. Some left for segregationist Strom Thurmond. Others defected to Henry Wallace, the founder and leader of the Progressive Party. With polls showing Truman trailing by about 13 points to Republican Thomas Dewey, many pollsters stopped surveying voters weeks before Election Day, assuming the outcome was obvious. And a day after the election, the Chicago Daily Tribune plastered the headline “DEWEY DEFEATS TRUMAN” across its front page.
However, in a stunning turn of events, Truman won the election decisively, outperforming the polls by around 17 percent.
While 1948 is an extreme case of polling error, it still highlights that regardless of how clear a race seems, it is impossible to predict the result confidently. Pre-election polls are never entirely accurate, and error is always expected.
Many people, particularly nervous Democrats, believe that current polls may again be underestimating Trump, given that he outperformed them in both 2016 and 2020. While polling errors in those elections indeed favored Trump, their magnitude was surprisingly small. In fact, the 2016 national polls were more accurate than those in 2012. Trump was also underestimated in 2020 (with one outlier poll giving Biden a 17-point lead in Wisconsin, a state he won by less than 1 percent), but the errors generally fell within the typical range seen in election polling. This discrepancy may even be attributed to Democrats exercising caution about COVID and avoiding crowded polling areas.
Some have speculated that these past polling errors were partly due to tendencies unique to Trump voters. In 2016, some evidence suggested that Trump supporters were likely to lie to pollsters because they felt that it was socially unacceptable to support him openly. While embarrassed Trump voters may have impacted the 2016 polls, this will likely not be the case in 2024. Trump has since become a well-established figure, making support for him less socially taboo. Furthermore, following his assassination attempt in July, his approval rating has surged to its highest point in years.
Another challenge is that Trump voters tend to distrust institutions, including polling firms, making them harder to reach in surveys. Pollsters are mindful of this fact and have adjusted their sampling methods and response weighting to better capture Trump’s base. These efforts have shown improvement: 2022 was a historically accurate year for political polling.
Polling averages are therefore the best information available for predicting the outcome of the election. (Note: all 2024 polling numbers I discuss from here to the end are from Nov. 2). Currently, all swing states are a typical polling error away from tipping to either candidate — the least close being Georgia, where Trump leads by about 2.5 percent. With margins this tight, even a small systematic polling error could swing all of them to one candidate, making the outcome of the election appear deceptively lopsided. Prediction markets give both Trump and Harris about a 20 percent chance of a sweep.
Among the swing states, Harris has polled the strongest in Michigan, Wisconsin, and Pennsylvania. Michigan, her best-performing state of the three, shows her currently ahead by 1.2 percent and she has not trailed there since July. Although Pennsylvania and Wisconsin remain essentially tied, Harris has generally held a narrow lead in both states for the past months. These three states each have a high percentage of white voters and share similar economic concerns, and, as a result, trend the same way electorally. Like Biden, Harris has polled better with white voters than Clinton did in 2016, helping her maintain a lead in these states. One potential wrinkle, however, is Michigan’s high concentration of Arab-American voters, who strongly disapprove of the Biden-Harris administration’s handling of the Israel-Hamas war.
Across the aisle, Trump has polled the strongest in Arizona, Georgia, and North Carolina. These three states have relatively large concentrations of nonwhite voters, among whom Trump has been performing better than any Republican in decades. Notably, Georgia, which voted for Biden in 2020 and elected two Democratic senators shortly after, is Trump’s strongest swing state this year.
Using polling data and weighting based on demographics and “fundamentals,” statisticians have been working to forecast the election. The three most respected models — Nate Silver’s Silver Bulletin, FiveThirtyEight, and The Economist — each give Harris and Trump equal odds of winning, with Pennsylvania projected as the likely tipping-point state. Prediction markets like Polymarket and Kalshi see Trump as an ever-so-slight favorite but agree that Pennsylvania will tip the election. Historian Alan Lichtman, famed for correctly predicting elections based on 13 (somewhat dogmatic) true-or-false questions, believes that Harris will win.
Both candidates have reasons for optimism. Trump can be encouraged by the trend of incumbents faltering worldwide in 2024. He might also hope that left-leaning independents, such as Jill Stein, will siphon votes from Harris, especially since right-leaning independent RFK Jr. dropped out and endorsed him. Additionally, Trump may be bolstered by Harris’s potential error of not choosing Josh Shapiro, the highly popular governor of Pennsylvania, as her running mate.
On the other hand, Harris may be heartened by the possibility that pollsters have added too much pro-Trump weighting in an effort to correct for previous underestimations. A recent high-quality poll showed Harris leading in Iowa, which has recently voted solidly red, suggesting that she may be underestimated in the Midwest. Additionally, Harris has a higher approval rating than Trump, which should translate to more votes. Furthermore, left-leaning mainstream media sources may be overestimating Trump’s chances due to negativity bias, perceiving a Trump victory as more likely because it’s viewed as an unfavorable outcome.
With all of these projections in mind, I will try to predict the election myself. I am writing this on Saturday, but given that polling has been relatively steady recently, there’s little reason to expect any major shifts by Tuesday. Also, many Americans have already voted, limiting the extent to which new political developments can influence the outcome of the election. It is worth noting that the early vote is not predictive of the election outcome — for example, in 2016, Clinton led the early vote in West Virginia by 12 percent, despite ultimately losing the state by over 40 points.
My prediction is that Trump will ultimately win the election. There will be no surprises in the deep-red and deep-blue states, leaving the election to be decided by Arizona, Nevada, Georgia, North Carolina, Pennsylvania, Michigan, Wisconsin and possibly Nebraska’s 2nd Congressional District, which votes separately from the rest of the state. I anticipate that Trump will carry Arizona, Georgia and North Carolina, while Harris will win Michigan, Nevada (by less than 0.5 percent) and Nebraska’s 2nd District, and eke out Wisconsin. Trump will win Pennsylvania — the tipping point state — by 0.3 to 0.8 percent, giving him an Electoral College victory of 281 to 257, despite Harris winning the popular vote by around 0.7 percent. The polls, I believe, will underestimate Trump by 0.5 to 1 percent.
This is just one story of how the election will unfold, and I will probably end up proven wrong in some regard. Either candidate could be underestimated in the same way Harry Truman was in 1948. Anything could happen on Tuesday — we’ll just have to wait and see.
Ben Aizenberg can be reached at benjamin_aizenberg@brown.edu. Please send responses to this op-ed to letters@browndailyherald.com and other op-eds to opinions@browndailyherald.com.
