Midterm 2026 Forecast Analysis
About the margin of inefficiency and how to read this page
This view replays the 2024 cycle through the same lens as the live 2026 board: every state presidential market and every Senate race market, the market consensus and its margin of inefficiency, and the forecasters measured against it, one model date at a time. Drag the date to see how the gap between the models and the markets evolved; the scorecard checks both against the results.
Markets: Kalshi's state presidential markets (listed 7 October 2024, prices from the full trade tape in our archive) and Polymarket's state presidential and Senate winner markets (daily prices from spring 2024). Neither venue's 2024 order books survive, so a quote is reconstructed as the day's closing price plus and minus half a cent, and Polymarket is treated as fee-free, as it was in 2024. Forecasts: Silver Bulletin's presidential model (every daily run from 29 July) and 538's presidential, Senate and chamber models (archived daily series). Model dates run to 5 November 2024.
I proposed the Margin of Inefficiency (MOI) (The Super Model, October 2024) during the 2024 election as a way to interpret the forecast implied by prediction market prices. The MOI accounts for explainable differences between different market prices that may be caused by access restrictions, platform fees, and other sources of market friction. The margin of inefficiency applies to estimates derived from prediction market prices in an analogous way to the margin of error in traditional polling estimates.
Below is a live display of the implied forecasts for the 2026 midterms based on live quotes from Kalshi, Polymarket, and other political markets. The market consensus is derived by combining the forecasts across these markets and accounting for the estimated margin of inefficiency for each market. Model forecasts are also shown for comparison to traditional poll aggregation models and polling estimates. An asterisk marks a market whose quotes cannot all hold at once; the figure shown is the point the quotes meet at after adding the smallest exogenous margin of inefficiency that makes them overlap.
Background: the original post, Margin of Inefficiency in Prediction Markets (The Super Model, October 2024), and the working paper, The Margin of Inefficiency: Reading Prediction Market Prices as Bounds on Probability (PDF, October 2026).
Quotes: Kalshi, Polymarket, PredictIt. Model forecasts: Decision Desk HQ, 51+1 (G. Elliott Morris), VotePredictor, Silver Bulletin (FLIPR). Handicapper ratings via Wikipedia.
Quotes: Kalshi (trade archive), Polymarket (price history). Model forecasts: Silver Bulletin 2024 presidential model, 538 2024 forecasts (Wayback Machine archive). Results: Kalshi settlements and Polymarket resolutions.