How the 2026 midterm forecast works.
This forecast projects the outcome of every Senate, gubernatorial, and House race in the 2026 midterm elections. It combines national polling, state-level surveys, and historical voting patterns into a probabilistic model that runs thousands of election simulations daily.
The model ingests three main signals: a generic ballot average computed from a quality-filtered, weighted rolling window of national polls; state and district polls where available; and historical partisan ratios derived from recent election results. These are blended together with weights that reflect their relative informativeness — state polls carry the most influence where they exist, and the generic ballot fills in everywhere else.
A national indicator is also computed by reverse-engineering the implied national environment from all available state polls, providing an additional cross-check on the generic ballot signal.
The model applies several corrections to improve accuracy. A pollster quality filter screens polls by historical track record and weights them accordingly. A circuit breaker in non-competitive states prevents the national trend from overriding strong local polling. A Hispanic voter adjustmentaccounts for shifts in Hispanic voter preference relative to recent baselines, scaled by each district's demographic composition.
Individual race win probabilities are derived from projected margins using a normal distribution that accounts for polling uncertainty. Chamber-level outcomes are determined through Monte Carlo simulation — each run draws a correlated national swing and resolves every race, producing a full election map. The share of simulations in which each party reaches a majority determines the control probability.
Nowcast reflects what would happen if the election were held today — a pure snapshot of current polling.
Forecast projects forward to Election Day by modeling how the race is likely to evolve. Undecided voters are gradually allocated based on historical patterns, and a small structural adjustment is applied to account for the typical relationship between midterm polling and final results. Both adjustments ramp gradually to full strength by early fall.
Each race is classified on a seven-point scale from Safe D to Safe R based on its projected margin. The thresholds are calibrated to reflect meaningful differences in competitiveness.
| Rating | Margin | Description |
|---|---|---|
| Safe D | >15 pp | Strong Democratic advantage |
| Likely D | 7.5–15 pp | Clear Democratic lead |
| Lean D | 2.5–7.5 pp | Slight Democratic edge |
| Tossup | <2.5 pp | Either party could win |
| Lean R | 2.5–7.5 pp | Slight Republican edge |
| Likely R | 7.5–15 pp | Clear Republican lead |
| Safe R | >15 pp | Strong Republican advantage |
National and state polls are sourced from public polling databases and manually curated releases. Election baselines come from recent presidential and congressional results. Demographic adjustments use Census population estimates. Maps use US Census Bureau shapefiles.