—
Results
The analysis
Seats by party. The Senate bar is the full 100-seat chamber, with seats not up this year counted for the party holding them.
Swing in two-party margin vs the state's 2024 presidential result. Click to switch office.
District's 2024 presidential margin → 2026 House margin (two-party). Sorted by size of swing; called flips only.
House districts, swing vs 2024 presidential two-party margin. Dashed line = median (the "uniform" national swing).
Each dot is a district. The 45° line is "no change"; the offset line is a uniform swing at the national median. Distance from the offset line is the local residual.
Two-party swing against the 2024 presidential vote, and against the last election for the same seat ("prior race").
| Office | Baseline | Median | Mean | SD | n |
|---|
Median district swing within each band of 2024 presidential margin; dashed line = all districts. Near-equal bars mean the wave hit red and blue turf alike.
District swing minus national median swing. Top and bottom 12 called races.
Senate and governor races with a published final aggregate (RealClearPolling, Silver Bulletin, Race to the WH). Above the line: polls understated Democrats.
Positive = polls understated Democrats. Prior cycles are approximate published cycle-level biases for Senate and governor polls; 2026 is this analysis.
Two-party margin points. CI is a 4,000-draw bootstrap of the mean.
| Office | n | Signed error | CI low | CI high | Abs error | RMSE | Understating D |
|---|
Actual minus final average. Bars right of zero: Democrats beat their polls.
—
House districts on unchanged 2024 lines only (states that redrew for 2026 are excluded because district demographics are on the old lines). Each panel: swing vs 2024 presidential two-party margin against one district characteristic, with an OLS fit.
Outcome: swing vs 2024 presidential, in two-party points. Coefficients are per one-unit change (income per $1,000); robust standard errors. Bold = p < 0.05.
| Level | Model | Term | Coef | SE | p |
|---|
Share of the variation in swing each model explains. Even the fullest district model leaves three-quarters unexplained.
| Level | Model | n | R² | Adj. R² |
|---|
Called races only. Dots are the implied probability each bucket conventionally carries; bars are the observed Democratic win rate. Perfect calibration = bar meets dot.
Democratic seats: final published forecasts (grey), two uniform-swing rules (light blue) and the count so far (blue).
President's party net House seat change at each midterm since 1994 against approval at the election. 2026 uses exit-poll approval and seats called plus leads.
Called races where the winner was rated as the underdog by Cook (state races) or DDHQ (House).
| Race | Office | Rating | Winner | Margin |
|---|
Two-party margins in points; swing vs the 2024 presidential result in the same geography. Click a header to sort.
| District | Democrat | Republican | D% | R% | Margin | Status | Flip | Pres '24 | Swing | DDHQ | Cook | % in |
|---|
Results. Decision Desk HQ race pages, read on Nov 4, 2026 (state races early morning, House around midday ET). A race is "called" only when DDHQ has made a call; everything else is shown as a lead. Percentages reported as ">95%" are stored as 95.
Baselines. 2024 presidential results by state (FEC) and by congressional district (Daily Kos Elections); prior results for the same seat from Wikipedia. States that redrew House maps for 2026 use presidential figures re-aggregated to the new lines where available.
Polling. Final published averages (RealClearPolling, Silver Bulletin, Race to the WH, via Wikipedia aggregation tables). Races without an established aggregate are excluded from polling-error statistics.
Ratings and forecasts. Final Cook Political Report, Sabato's Crystal Ball and Inside Elections ratings (Oct 29–Nov 2), DDHQ's own House categories, and published seat projections from Silver Bulletin and DDHQ.
Demographics. ACS 2018–22 (education, income) and 2020 Census (race, urbanization) by state and by 118th-Congress district; FiveThirtyEight urbanization index.
Margin. Two-party: (D − R)/(D + R) × 100, so third-party and top-two same-party contests don't distort swing.
Swing. 2026 margin minus the baseline margin. "Uniform swing" is the median across races; the local residual is each race's swing minus that median.
Polling error. Actual margin minus final-average margin; positive means polls understated Democrats. Confidence intervals are percentile bootstraps of the mean.
Calibration. Ratings mapped to conventional implied probabilities (Solid .98, Likely .85, Lean .70, Toss-up .50). The Brier score compares these with outcomes; climatology is the score from predicting the overall D win rate for every race.
Caveats. Counts are preliminary and will move in late-counting states (CA, AZ, WA, OR, NV, AK). Regression coefficients describe association, not cause; district-level demographic models exclude the nine states that redrew their maps.