In states Trump carried, the sources take anywhere from 0 to 6 points off the Democratic poll lead
UnclassifiedMost of the seat-level disagreement is how far to mark down the Democrat's poll margin in a Republican state. In South Carolina Kalshi rose from 20% to 24% since October 25 and now takes 2.5 points off, against FiftyPlusOne's 0.5 at 34.2%. In Ohio FiftyPlusOne fell to 62.7% and takes off 1.7, Kalshi 0.9 at 67.5% and DDHQ 3.6 at 51%.
A markdown of 2 to 5 points is what the record supports, and the market's sits in that band in South Carolina, Ohio, Kansas and Iowa while FiftyPlusOne's sits under it in all four. The exception is Texas, where the market prices the poll lead nearly at face value; there DDHQ's 1.8-point markdown is closer to the record.
Why these should match, and what could explain the gap
Every source reads the same polls. Turning each probability back into a margin shows what it does to them. The market takes 2.5 points off the Democrat in South Carolina and Kansas, 1.3 in Iowa and 0.9 in Ohio, adds 0.1 in Texas, and adds 1.7 to Peltola's lead in Alaska, where its price is 74.5% against a poll average of Peltola +2.6. DDHQ takes off 5.6 in Kansas, 3.6 in Ohio, 1.7 in South Carolina, 1.3 in Texas and 1 in Alaska, and adds 0.4 in Iowa. FiftyPlusOne takes off 1.7 in Ohio, 1 in Kansas, 0.9 in Texas and 0.5 in South Carolina, and adds 0.3 in Iowa and 0.1 in Alaska. South Carolina moved most: Kalshi went from 20.5% on October 24 and 20% on October 25 to 24% today, with the poll average unchanged at Graham +2.1, so its markdown went from 3.3 points to 2.5. FiftyPlusOne's October 26 reading has Ohio at 62.7% from 64.5% and Kansas at 48.7% from 51.2%, against poll averages of Brown +3.8 and Hamilton +0.8. Kalshi is at 67.5% in Ohio, 64.5% in Texas (poll average Talarico +2.3) and 41.5% in Iowa (Hinson +0.1). Against the 2-to-5-point band in the record, the market sits inside it in South Carolina and Kansas and under it in Iowa, Ohio and Texas; the October 3 call rested on four seats inside the band and today there are two. DDHQ sits inside it in Ohio, under it in South Carolina, Texas and Iowa and above it in Kansas; FiftyPlusOne sits under it in all six. DDHQ's numbers are dated October 2.
- The record: Republican-state Senate polls overstated the Democrat in 17 of 19 midterm races since 2014, by 5.3 points on average, and by 2 to 3 points in the contested races of 2018 and 2022.
- Base rates: out-party challengers have beaten an incumbent of the president's party in a state this red in 2 of 37 midterm races since 2006; open seats went 2 of 8.
- Alaska has votes behind its premium: Peltola out-polled Sullivan by 8 in the August primary, and the newest poll (New York Times/Siena, September 24 to October 1) has her up 50 to 43. Its Senate polls missed the other way in 2020, understating Sullivan's margin by about 8 points. Texas has only polls, and its final averages overstated the Democrat by 3 to 4.5 in the last two cycles.
- A third model, read once. Silver Bulletin's Deluxe forecast (October 4, subscriber page, not a board source) publishes projected margins, so its markdown can be read directly: 2.7 points off its own poll average in South Carolina and 2.1 in Kansas, inside the 2-to-5-point band; 0.4 off in Ohio, nothing in Iowa and 0.3 added in Texas, under it. Its South Carolina number is 19.8%, nearer the market's 14.5% than FiftyPlusOne's 36.1% that day. In Alaska it has Peltola at 62.9% on a 3.1-point first-choice lead, between the other two models and 11 points under the market; two other Republicans hold 4.2 points of first-choice vote in its projection, which under ranked-choice counting may be why the lead converts to a lower probability there than in any other tracked seat.