Forecasting the 2024 U.S. General Election

Abstract

This paper forecasts the outcome of the 2024 U.S. general election. Using the poll of polls approach, a Bayesian hierarchical model was developed with data from FiveThirtyEight using polling data from July 21 to October 19, 2024. The model’s prediction is based on the electoral college vote in the seven key swing states: Arizona, Nevada, Michigan, Georgia, Wisconsin, Pennsylvania, and North Carolina to estimate the overall winner. Results indicate that Harris is predicted to win with a probability of 66%.

File Structure

The repo is structured as:

• data/raw_data contains the raw data as obtained from FiveThirtyEight on October 17th.
• data/analysis_data contains the cleaned dataset that was constructed for analysis.
• model contains the fitted models.
• other contains relevant literature, details about LLM chat interactions, and sketches.
• paper contains the files used to generate the paper, including the Quarto document and reference bibliography file, as well as the PDF of the paper.
• scripts contains the R scripts used to simulate, download and clean data.

Statement on LLM usage

Some parts of the code and writing were created using the auto-complete tool, ChatGPT. ChatGPT-4 is used when writing the abstract, introduction, appendix, and data simulation code, with the full chat history documented in inputs/llms/usage.txt.