Model-based race ratings.
Ratings combine historical results, polling where available, and structural factors such as incumbency, retirements, and national environmental factors.
CM Election Predictions tracks U.S. elections with model-based ratings and live election results. The goal is simple: make the electoral landscape easier to read, and make results easier to understand.
Ratings combine historical results, polling where available, and structural factors such as incumbency, retirements, and national environmental factors.
Election night tools are designed to give users the ability to easily understand results as they come in, and what they mean for wider results.
Forecasts are meant to be evaluated over time. Past projections and actual outcomes are tracked on the Historical Results page.
The Model incorporates historical election results, available polling data, and adjustments for national environmental factors like incumbency, retirements, and election cycles (midterm vs. presidential). It estimates likelihoods, not certainties.
Inputs are applied consistently. Ratings are not manually overridden to match conventional wisdom, even when polling data is sparse.
You can view the full performance record, including past projections and actual outcomes, on the Historical Results page.
Early in a cycle, polling data may be limited or unavailable. During that period, the Model relies more heavily on historical results and structural factors, which can produce ratings that look different from conventional expectations.
As polling volume increases, the forecast updates to reflect the new information. States with unusual recent voting histories may remain more competitive in the model than someone may assume from recent elections.
Prediction pages update around every 60 minutes.
CM Election Predictions is an independent election forecasting project created by Collen Mahoney. It is built to apply consistent methods to election forecasting and make the data easier to understand.