For the complete documentation index, see llms.txt.
Skip to main content
New
20% · 1/5
Lesson 12 min 20 XP

FiveThirtyEight's Methodology

A deep dive into how the most famous election forecast model works: pollster ratings, state correlations, and uncertainty modeling.

Inside the Model

FiveThirtyEight's model, developed by Nate Silver, is built on several key components. First, a polling average for each state, weighted by poll quality (rated A+ through D), recency, and sample size. Second, a fundamentals component that uses economic indicators and presidential approval to set a baseline prediction before polls become available. Third, a state correlation matrix that models how states move together: if the model sees a shift in Michigan, it infers a likely shift in Wisconsin and Pennsylvania.

The model also incorporates 'house effects,' systematic biases in individual pollsters. If Rasmussen consistently shows Republicans 2 points higher than other pollsters, its results are adjusted accordingly. This does not assume Rasmussen is wrong; it ensures that variation between pollsters does not distort the average.