2024 US presidential election

Updated 2024-11-05 08:46 CET

Harris has a 50.0 % chance of winning. Method

If polls were 100% accurate, results right now would be Harris 273, Trump 265. Odds of a tie: 0.7 %.

State Polls EVs Avg Odds
Nebraska CD-331-51.3 %0.0 %
Wyoming53-41.4 %0.0 %
West Virginia55-39.1 %0.0 %
Oklahoma87-33.5 %0.0 %
North Dakota63-31.3 %0.1 %
Idaho34-31.2 %0.2 %
South Dakota53-27.0 %0.4 %
Kentucky58-27.4 %0.5 %
Utah106-22.6 %0.6 %
Alabama59-26.4 %0.7 %
Arkansas56-25.5 %0.8 %
Tennessee1011-23.4 %1.0 %
Montana103-18.5 %1.8 %
Nebraska102-15.5 %3.0 %
Louisiana58-18.7 %3.9 %
Mississippi46-17.1 %5.3 %
Indiana511-16.8 %5.4 %
Missouri1010-13.6 %6.5 %
Nebraska CD-131-15.1 %6.7 %
Kansas56-14.1 %8.4 %
South Carolina109-11.9 %9.0 %
Ohio1018-7.6 %10.6 %
Florida1029-6.1 %13.3 %
Texas1038-6.2 %15.4 %
Alaska103-8.9 %17.0 %
Maine CD-2101-5.2 %27.0 %
Arizona1011-3.2 %28.4 %
Iowa106-4.0 %32.5 %
North Carolina1015-1.4 %39.4 %
Georgia1016-1.4 %39.6 %
Nevada106-1.0 %42.4 %
Pennsylvania10200.1 %50.6 %
Wisconsin10101.1 %58.2 %
Michigan10161.1 %58.4 %
Minnesota10105.1 %78.8 %
New Mexico1056.7 %79.8 %
Maine1029.0 %84.8 %
Nebraska CD-21019.4 %85.3 %
Virginia10136.7 %85.9 %
Oregon6712.6 %88.7 %
Colorado10912.4 %91.7 %
New Hampshire1049.1 %92.6 %
Delaware5315.8 %93.7 %
Illinois52016.4 %93.9 %
New Jersey71414.2 %94.2 %
Connecticut5716.6 %94.2 %
Rhode Island10414.7 %94.5 %
New York102917.0 %98.0 %
Washington101218.3 %98.2 %
Maine CD-19122.8 %99.5 %
Hawaii4430.0 %99.7 %
Vermont6331.9 %99.9 %
California105523.9 %100.0 %
Massachusetts101127.5 %100.0 %
Maryland101027.9 %100.0 %
District of Columbia2386.8 %100.0 %
 Blank map of the United States, territories not included Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming District of Columbia District of Columbia

Development over time

Produced by running the model and for each date ignoring all polls completed after that date. Because there is a lag between polls being completed and results coming out, the curve tends to dip at the end. Don't read too much into the last 3-4 days.

The red line is the Biden/Trump debate disaster, and the black line is when Biden withdrew.

2024-11-04T16:03:10.768698 image/svg+xml Matplotlib v3.9.0, https://matplotlib.org/

How results are computed

The starting point is the set of all polls from fivethirtyeight.com

From the 538 data, a poll average for all states is computed. Only the 10 latest polls are used, and the polls are weighted by how old they are, and how many respondents there are. Polls marked as partisan by 538 get lower weight.

Because some safe states have very poor polling the election results from 2016 and 2020 are included. Those are only used for states with few polls, and they are weighted lower than recent polls.

Then 100,000 simulations are run. For each simulation, a national poll bias is picked from a normal distribution. All averages have this number (usually -5 to +5) added to them. So if Michigan was at +2, but the national bias is -5, Michigan is now at -3 for this simulation.

Finally, a local bias is computed for each state. The normal distribution is sized based on the sum of the time-weights for the polls we have for that state. If we only have one old poll the sum is 0.1, and the distribution is wide, because we are uncertain. If we have 10 recent polls the sum will be something like 7, and the distribution will be narrow, because we are pretty sure of the results.

Let us say Michigan gets a +2.7 local bias. Then it ends up at -0.3. Each state is called for whoever this twice-adjusted average favours. In this case, Trump.

This simulation is run 100,000 times, and the win probabilities for each state and overall are calculated from that.

Author: Lars Marius Garshol.