New Alaska survey splits within a point; our forecast does not move

An Alaska Survey Research poll of 1,495 likely voters, fielded August 20-23, put Mary Peltola at 50.6 percent and Dan Sullivan at 49.4 percent. That 1.2-point gap is small enough to describe the race only as close.

Our forecast did not respond. It gives Peltola a 24 in 100 chance and Sullivan 76 percent, with a mean margin of R+8.2 and a 5th-to-95th-percentile range running from R+26.6 to D+10.3. The run before it, from 12:32 PM ET on September 2, had the same win probabilities, a mean margin of R+8.1, and a range of R+26.7 to D+10.4. The change is inside the noise of the simulation.

The gap between the survey numbers and our margin is the durable feature of this race in our model. Poll results showing Peltola at or ahead of Sullivan sit alongside a projected mean of Sullivan by roughly eight points, and the model's stated range still spans outcomes from a Sullivan win by more than 25 points to a Peltola win by about 10. That width, not the point estimate, is what a 24 percent chance is built on.

Nationally, our model gives Democrats a 48 in 100 chance of Senate control and Republicans 51 percent, with a mean of 50.2 Democratic seats. Democrats need 51 seats for control; Republicans need 50, because the vice president is a Republican and a 50-50 tie goes to them. Alaska is one of the seats that decides which side of that line the chamber falls on. In the House, our model puts Democratic control at 93 percent against 6 percent for Republicans, with a mean of 240.1 seats against the 218 needed. Our House error model correlates districts nationally, by census division and by state, but has no term linking demographically similar districts in different places, so that 93 percent is marginally firmer than a fuller accounting would leave it.