Cook Political‘s House ratings update on 25 September 2026 moved 15 races in Democrats’ favour, and the Cook Political House ratings revision is being read by most outlets as confirmation that Republicans are headed for a bruising November. That read is broadly correct. The part worth pressing on is how much further the map may still move, and whether the current projections are already being overtaken by the underlying data.
What the Ratings Actually Show
After the 15-race shift, Cook now rates Democrats as favoured in 208 seats and Republicans as favoured in 205. Democrats currently hold 215 seats and need a net gain of just three to retake the majority. Cook’s central projection, assuming Democrats win every toss-up, puts them at 220 to 230 seats, a workable but narrow majority by modern standards.
The toss-up category is where the consensus narrative may be moving too slowly. Axios reports that 17 Republican-held seats are now rated as toss-ups, compared with just five Democratic-held seats. That asymmetry does not fully register in summary headlines about a “slim Democratic majority.” A 17-to-5 toss-up imbalance, even if Republicans split those races roughly evenly, still points toward a Democratic gain well above the three seats required for control.
Among the seats reclassified in this round, The Washington Post notes that three races, including Rep. Monica De La Cruz’s seat in Texas, moved from likely Republican to toss-up. Texas seats appearing in this category at all is the detail worth sitting with. The state was not competitive House territory in recent cycles, and its appearance here reflects something broader about where Republican structural weaknesses are spreading.
The Cook Political House Ratings and the Undecided Voter Question
Cook’s own commentary points to a plausible Republican escape route: undecided voters in lean and likely Republican districts have historically broken toward the GOP, and in districts that backed Trump by double digits in 2024, that tendency should hold. Republican operatives are banking on it.
The problem with that argument, at this particular moment, is that it relies on base behaviour returning to normal in an environment that is not normal. Cook acknowledges the president is deeply underwater on the economy and is dragging down House candidates across competitive districts. If presidential approval remains where it is through late October, the undecided-voter safety valve may not release enough pressure to matter.
That is also roughly the implied reading from the forecasting models sitting alongside Cook’s projection. Nate Silver’s Deluxe model projects 237 Democratic seats, seven to 17 above Cook’s range, with the probability of a Democratic House majority at 68.6% under the Deluxe model, 65.9% under the Classic, and 70.1% under the Lite. Kalshi’s prediction market sits at 62%. Cook does not publish an explicit probability, but its commentary, read carefully, implies something in the 60% range for Democratic control of the House, which puts it at the lower end of where the markets and Silver are priced.
The gap between Cook and Silver is not trivial. A projection of 230 seats versus 237 may look like a rounding difference, but it reflects different assumptions about how far a Democratic wave could penetrate into territory that was considered safe Republican ground even six months ago. The toss-up count, and the geographic spread of races like De La Cruz’s in Texas, suggests Silver’s end of the range deserves more weight than the headline Cook number implies.
Cook notes, correctly, that television advertising only started in earnest this month and that ratings will continue to shift. The direction of those shifts, given that all 15 in this round moved the same way, is not a coin flip. With six weeks to go and the competitive map tilted 17-to-5 in Republicans’ exposed seats, the question is less whether Democrats take the House and more whether the final margin lands closer to Cook’s conservative end or Silver’s more expansive one. The next round of Cook Political House ratings changes will clarify that quickly.
