Morris Says Race Raters Underestimate Democratic House Odds by About 30 Points

Two House pictures at 44 days

On Thursday G. Elliott Morris published a calibration study of the three major race raters as of about 45 days before Election Day. Read literally, Cook Political Report sat near 206 seats at least leaning Democratic, 208 at least leaning Republican, and 21 toss ups; Inside Elections near 200 Democratic, 202 Republican, and 33 toss ups; Sabato's Crystal Ball near 210 Democratic, 203 Republican, and 22 toss ups (G. Elliott Morris, September 18, 2026). Average those literal boards and you get about 218 Democratic seats, an exact toss up for control.

Morris's historically debiased reading of the same boards moves the three rater average to about 228 Democratic seats and roughly an 83 percent chance of a Democratic House majority (G. Elliott Morris, September 18, 2026). That is the gap Sunday's desk is scoring: not a cheer for either party, and not a certified Speaker. It is the difference between treating a toss up label as a fair coin and treating it as a label that has historically broken toward the out party in midterms.

What the historical midterm record shows

Morris subset race ratings to midterm years as of 45 days before Election Day and recoded every label as favoring the president's party or the opposition. Across Cook's House record in 2010, 2014, 2018, and 2022, the out party won 67 percent of seats labeled toss up. Lean seats favoring the president's party held about 78 percent of the time, while lean seats favoring the opposition held about 95 percent of the time (G. Elliott Morris, September 18, 2026). Inside Elections and Sabato show the same directional pattern: all three give the out party about two thirds of toss ups.

The 2018 example is concrete. Forty five days before that midterm, Cook rated 28 Republican held seats as toss ups; Democrats won 23 of them, an 82 percent hit rate (G. Elliott Morris, September 18, 2026). That is not a claim that 2026 must copy 2018. It is a claim that a fifty fifty reading of today's toss up column is not how those labels have behaved when the White House party is on defense.

Debiased boards versus markets and models

Under Morris's empirical translation, Cook's implied Democratic seat total rises from about 217 to 224 with roughly a 75 percent majority probability; Inside Elections from 217 and 49 percent to 230 and 87 percent; Sabato from 220 and 57 percent to 231 and 89 percent (G. Elliott Morris, September 18, 2026). Pool the three and the House moves from a coin flip near 218 seats and 51 percent to about 228 seats and 83 percent.

Other public instruments already sit nearer the debiased side. Saturday wraps put Kalshi's Democratic House control price near 90 percent (New York Post, September 19, 2026). Morris notes that FiftyPlusOne's model has Democrats at a median of 238 House seats and a 98 percent chance of the majority (G. Elliott Morris, September 18, 2026). US Political Rank publishes the rater boards, the market prices, and the model claims side by side on most competitive House districts of 2026 and what history says about the 2026 midterm House. Readers can watch which instrument ages better.

Senate boards still disagree with House markets

Morris saved a full Senate calibration for a later piece, but the live Senate instruments remain split. Cook lists seven Senate toss ups after the September 15 New Hampshire move: Alaska, Iowa, Maine, Michigan, New Hampshire, Ohio, and Texas, with Georgia and North Carolina as lean Democrat (Cook Political Report, September 15, 2026; Newsweek, September 18, 2026). Sabato still showed a narrow Republican edge near 48 to 47 in Saturday wraps even as Kalshi priced Democratic Senate control near 58 percent (New York Post, September 19, 2026).

Democrats still need a net gain of four Senate seats because Republicans hold 53 to 47 and Vice President JD Vance would break a 50 to 50 tie. A House market near 90 percent Democratic and a Senate board that still gives Republicans a slight edge can both be true on the same Sunday. Our ranking of 2026 Senate races keeps every seat on one ruler while those instruments move.

What the calibration settles and what it does not

Settled on Sunday's desk: the three major raters still publish House boards that read as a coin flip if toss ups are treated as fair coins; Morris's historical midterm calibration moves that picture to about 228 seats and roughly 83 percent Democratic; out parties have historically taken about two thirds of Cook toss ups at the 45 day mark; Kalshi and FiftyPlusOne sit closer to the debiased House number than to the literal coin flip (G. Elliott Morris, September 18, 2026; New York Post, September 19, 2026).

Still open: whether 2026 breaks the historical midterm pattern; whether deep red defensive spending and redistricting change the toss up break rate; whether Senate calibration matches the House finding; and whether October polls force the raters to move before early vote share hardens. Handicappers still see candidate quality and local data that models miss. Models still catch national tides that labels lag. Score both. Leave the gavels to November 3.