Can a simple median-based polling method predict the next U.S. president?

In the high-stakes arena of U.S. presidential elections, swing states often decide the outcome – but how reliable are traditional polling averages? This study explores an alternative approach: the state-level median polling method, which sidesteps outliers and noise by focusing solely on the middle poll in each battleground state. Applying this technique to October 2024 data from seven critical swing states, we uncover surprising insights – including three accurate Trump projections, a miss in Michigan, and a dead heat in three others.

Did this method hint at Trump’s eventual victory before the votes were cast? And could refining it with only the most authoritative polls make future predictions even sharper? Dive into the analysis to see how medians, not averages, might hold the key to forecasting America’s most unpredictable elections.

The methodology for predicting the winner of the U.S. presidential elections using state-level median polls is as follows:

1. Identifying Swing States

First, determine which states are considered "swing states" – those where the race between the two major candidates is extremely close and could favor either side.

2. Collecting Polling Data:

Gather the most recent and credible polls from each swing state, reflecting voters' current preferences.

3. Finding the Median for Each State:

3.1. Arrange polls from most to least favorable for each candidate.

3.2. The middle poll is the median. If there are two medians, take their arithmetic mean. It is important to note that the median, not the average of all polls, is used.

4. Awarding Electoral Votes Based on the Median:

The candidate with the higher median in a state is predicted to win its electoral votes.

5. Summing Electoral Votes:

Combine the projected electoral votes from swing states with those from non-swing states. The first candidate to reach 270 electoral votes is predicted as the winner.

2024 Election Prediction Using This Method

Using October 2024 polls (from September 27 to October 27), the following seven swing states were analyzed: Pennsylvania, Georgia, North Carolina, Michigan, Arizona, Wisconsin, and Nevada.

Pennsylvania

Georgia

North Carolina

Michigan

Arizona

Wisconsin

Nevada

FINAL

Results:

• Trump led in Pennsylvania, Georgia, and Arizona (46 electoral votes combined).

• Harris led in Michigan (15 electoral votes).

• No clear leader emerged in North Carolina, Wisconsin, and Nevada (32 electoral votes combined), as Trump and Harris had identical medians.

  
  

Given these results, Trump held a slight advantage, but no decisive victory margin was evident.

Prediction Accuracy:

The method correctly predicted Trump's wins in Pennsylvania, Georgia, and Arizona but inaccurately projected Michigan for Harris.

Poll Sources:

All swing-state polls were aggregated from:

• Polls | FiveThirtyEight

• Latest Polls 2024 | RealClearPolling

• 2024 Presidential Election Polls: Harris vs. Trump - 270toWin

• The New York Times. Election 2024 Polls: Harris vs. Trump

  
  

Each included poll in the tables above was verified through at least one of these four sources. A survey’s potential bias toward either party should also be considered. For greater accuracy, future analyses could prioritize the most authoritative polls before calculating the median.

Conclusion

The state-level median polling method offers a compelling alternative to traditional polling averages. By focusing on the median – rather than the mean – this approach mitigates the influence of extreme polls, delivering a more stable snapshot of voter sentiment in critical swing states.

Our analysis of the 2024 election demonstrated both the promise and the challenges of this method. It accurately predicted Trump’s victories in Pennsylvania, Georgia, and Arizona but fell short in Michigan.

While the median method provides a valuable tool for cutting through polling volatility, its accuracy hinges on poll quality, timing, and sample composition. Future refinements – such as weighting polls by historical reliability or adjusting for late-breaking trends – could sharpen its predictive edge.

One thing remains clear: in an era of polarized electorates and ever-shifting battlegrounds, no single model guarantees certainty. Yet by isolating the median, we move closer to a clearer, more resilient way of reading the electoral tea leaves. The quest for the perfect forecast continues, but this method proves that sometimes, the middle ground holds the truth.

July 17, 2025

Nikita Senushkin

BA in Regional Studies