Nate Silver's FLIPR Model Launches Full 2026 Midterm Forecast Covering Every House, Senate, and Governor Race

Nate Silver and Eli McKown-Dawson have launched FLIPR, Silver Bulletin's comprehensive 2026 midterm election forecasting model, covering every competitive Senate, House, and gubernatorial race in the country. The model runs 40,000 simulations to produce probabilities for each race and models all three chambers jointly — meaning it can calculate compound outcomes like Republicans holding the Senate while losing the House simultaneously.

The polling methodology underpinning FLIPR is notably sophisticated. It adjusts for whether surveys were conducted among registered versus likely voters, corrects for house effects (systematic partisan lean in individual pollsters), weights polls by historical reliability, and applies a 'timeline adjustment' that revises stale polls to reflect shifts in the generic congressional ballot. The generic ballot itself gets a likely-voter adjustment before feeding into the model. Poll data is drawn primarily from The New York Times poll database, while generic ballot and congressional approval data are collected independently.

One of the more consequential early findings: mid-decade redistricting in Southern states is tilting the map in Republicans' favor in a meaningful cluster of House races. Silver and McKown-Dawson flagged this dynamic in a separate tour of 48 key House contests, noting that Democrats have stronger pickup opportunities in other parts of the country but face structural headwinds in the South specifically because of new district lines drawn outside the traditional post-census redistricting cycle.

FLIPR is a significant methodological expansion from Silver's prior forecasting work, with the team noting there are substantial changes to how the 2026 model operates compared to previous cycles. The full methodology is published separately for those wanting to dig into the modeling choices. The forecast updates continuously as new polls arrive, and the landing page is optimized for web viewing given the density of interactive charts involved.