Decoding the 2024 Election: Inside My Statistical Forecast Model
These eleven key metrics (and a bunch of uncertainty) shape my forecast of the Biden-Trump rematch.
In presidential elections, polls and data play a crucial role. They’re our best window into how things might go on Tuesday, November 5, 2024. As a high school student with a passion for statistics, I've attempted to create a forecast model for the 2024 presidential race that looks beyond just polls and who is ahead any given day.
My model combines eleven different factors, from how states have voted in the past to how much money candidates are raising, aiming to offer a detailed prediction of this important election.
I invite you to follow along as the forecast comes together, thinking about what it might suggest about our country’s political landscape and the state of democracy.
1. State Partisan Lean
This metric reflects how a state voted in the last two elections compared to the national average. It is essential for comparing national vote trends and polling against state partisan lean, providing a measure of deviation. This metric is weighted at approximately 4% and will gradually decrease to around 2% as election night approaches.
2. Last Election
Similar to the state partisan lean, this metric focuses exclusively on the 2020 election. It combines the national polling average to estimate how the state should vote in the current election. This metric holds a weight of 13%, decreasing to about 7% by election night.
3. Experience Winning Elections
Typically, this value indicates which candidate is perceived as more professional based on their highest political office. In this election, since both candidates' highest office is the presidency, this metric is set to 0. It currently accounts for about 0.3% of the model, gradually decreasing to 0.2%.
4. Fundraising
This measures the total individual donations a candidate has received during the election cycle, reflecting voter enthusiasm and available campaign resources. It currently comprises 1% of the forecast, decreasing to 0.5% by election night.
5. National Environment
This metric evaluates head-to-head polling matchups between Trump and Biden, including third-party candidates. It primarily complements the last election grouping, accounting for 0.3% of the forecast, gradually decreasing to 0.1% by election night.
6. Polling Average
Polling averages are the cornerstone of the forecast, presently constituting around 61% of the forecast, and will increase to 78% by election night.
7. Candidate Quality
This metric assesses which candidate is favored to win based on the 13 Keys to the White House. Currently, Biden is favored in 9 out of 13 keys, with a projected margin of victory of 6.6 points. This metric holds a weight of 0.3%, decreasing to 0.1% by election night.
8. Approval Ratings
This metric combines statewide approval ratings for Biden with Trump’s approval, factoring in projected improvements for Biden based on historical trends of incumbent approval ratings. It currently holds 5% of the forecast and will remain relatively stable until election night.
9. Projected Partisan Lean
This metric projects the partisan lean of a state based on the last five presidential elections, using a best-fit graph. It currently constitutes 14% of the forecast, decreasing to 7% by election night.
10. Incumbency Bonus
Given the historical trend of incumbents performing better in re-election, Biden is given a bonus currently equivalent to 1.4 points, which will gradually decrease to 1 point by election night. This bonus remains due to the tendency of incumbents being underestimated in polls.
11. Expert Ratings
This metric uses projections from Sabato's Crystal Ball, Cook Political Report, and Inside Elections to determine the projected margin for each state.
For states with a partisan lean exceeding 20 and rated as solidly one-sided, the partisan lean is used instead of the projected margin. The combined expert ratings are weighted as 20% of the forecast, with the remaining 80% allocated to other factors.
Uncertainty Factors
Several factors increase uncertainty, including:
Distance from election night
Significant shifts in polls
High numbers of undecided voters
Quality and quantity of polls in a state
Use of ranked-choice voting (e.g., Alaska)
High partisan lean
Discrepancies between state polling and national trends
Deviations of the forecast from polling data
If the state is a congressional district
Sources
Campaign fundraising: Wikipedia
Past election results: Jackson Cooley Election Maps
General election polling and statewide approval polls: Race to the White House
Trump favorability polls: FiveThirtyEight
Candidate quality: 13 Keys Tracker