Election Prediction

2024 Presidential Election Prediction Model

By Adam Koplik

Overview

This project forecasts the outcome of the 2024 U.S. presidential election using a weighted Bayesian probability model applied to national and swing state polling data. The goal was to create an interactive, real-time dashboard that updates predictions for each state and the overall election outlook.

Data Sources

• National and state-level polling data from major aggregators
  
• 2020 election results for model calibration
  
• State electoral vote allocations

Methodology

• Built a weighted Bayesian model that updates candidate win probabilities as new polls come in.
• Incorporated weightings based on:
  • Pollster historical accuracy
    
  • Sample size
    
  • Recency of the poll
    
• Simulated 10,000 election scenarios to generate state-by-state probabilities and national outcome distributions.
• Developed an RMarkdown dashboard hosted via ShinyApps.io to visualize live election forecasts and swing state probabilities.

Key Features

• Swing State Win Probability Map
• State-by-state vote share predictions
• Projected Electoral College outcomes
• Interactive dashboard viewable online

Visualizations

Tools Used

• R, RStudio
  
• brms, tidyverse, shiny, ggplot2
  
• RMarkdown for dashboard construction
  
• ShinyApps.io for deployment

Live Dashboard

📊 View the interactive dashboard here

Limitations

• Model assumes polls are unbiased on average, though real-world polling error is inevitable.
• No adjustments made for early voting patterns or state-level ballot access complications.

---

Built for my Statistics Senior Seminar, Fall 2024.