Skip to content

535 Final forecast · Nov 3
Dispatches

More

← All dispatches

Movement note AK-1

Our Alaska at-large model now gives Democrat Eric Hafner a small chance after a week at 0%

Over seven days our model's Democratic win probability in AK-1 rose from 0% to 9%, while the Republican's chance rose from 36% to 91%, and no single input explains the shift.

Written by the newsroom · anthropic/claude-opus-5.5 · October 6, 2026 at 12:32 AM EDT

Correction: the forecasts in this article predate a polling-data repair. Some imported polls omitted candidates who shared a party, and ranked-choice rounds were not consistently selected. The current forecasts use corrected results and recalculated pollster adjustments. See current forecasts.

Our model now gives Democrat Eric Hafner a 9 in 100 chance of winning Alaska's at-large House seat. In the run of September 29, seven days earlier, it put his chance at 0%. That is a move of 9.1 points toward the Democrat.

The same interval brought a larger change elsewhere in the forecast. Our model put Republican Nick Begich III's chance at 36% on September 29 and at 91% now. The earlier run gave 64% to a candidate outside the two parties. The current run lists no separate figure for independent Bill Hill. Our average margin moved from D+3.1 to R+11.1. The current 5th-to-95th percentile range runs from R+25.0 to D+2.6, against R+11.2 to D+17.4 a week ago.

Three New York Times/Siena University ballot tests, fielded September 24 to October 1, finished after the earlier run. In one, a sample of 487 likely voters, the Democrat drew 8.59% to Begich's 48.87% and Hill's 36.78%. In another, a sample of 504 likely voters, the Democrat drew 8.19% to Begich's 46.59% and Hill's 35.06%. A third, a sample of 476 likely voters, tested only Begich and Hill, at 53.04% and 46.96%.

Those polls show the Democrat in single digits. Nothing in them accounts on its own for a rise in his win probability. Our model moved, and the payload for this run does not identify why. We are leaving the cause open until it can be traced.

Polls cited

  • The New York Times/Siena University · Oct 1, 2026 · n=476 · AK-1
  • The New York Times/Siena University · Oct 1, 2026 · n=487 · AK-1
  • The New York Times/Siena University · Oct 1, 2026 · n=504 · AK-1

Every constant behind these numbers is on the methodology page.