Forecasts of the 2026 midterm elections from polls, prediction markets, academic models and the PLSC 2219 class model, updated daily.
Forecasts of the 2026 U.S. House and Senate elections, collected every day. Each line on the chart is one type of source.
| Polling | Polling averages of the national House vote (the generic ballot) and of individual races. |
| Markets | Prices on Kalshi, Polymarket and PredictIt, read as probabilities. |
| Academic | Forecasting models from published political science research, run on current data. |
| Class model | The model built in PLSC 2219, re-run each day with that day's presidential approval, gas prices and real income. |
Each line is the average of the forecasts in it on that date. When a forecast is added or stops updating, the line can move even if no forecast changed.
Seat counts for polling and some academic models come from carrying the national vote to each seat using its partisan lean. For the polling line's Senate numbers, races with their own polling averages use those.
The collected forecasts will be published as a dataset after the election, with a full description of how each source was collected and processed.
Data collected by this site are released under CC BY 4.0, except where a source's own terms apply (for example, figures drawn from Wikipedia are CC BY-SA). Code is under the MIT License. If you use them, please cite the site.
The website and the code that collects and processes the data were written with Claude, an AI model made by Anthropic, working under my direction. That includes the scrapers, the data processing, the charts and the page templates.
| Author | Kevin DeLuca, Department of Political Science, Yale University |
| Course | PLSC 2219, Midterm Elections and Forecasting, Yale University |
| Election returns | MIT Election Data and Science Lab (CC0) |
| Economic data | Federal Reserve Bank of St. Louis (FRED) |