The bump came partially as a result of Sunday’s New York Times/Sienna poll, which Silver has as one of the highest-rated pollsters. Silver’s newest rating has Trump with a 63.8% probability, compared to Harris’s 36%. The former president is also favored to win every swing state.

Silver’s current odds give Trump a 64% chance of winning Pennsylvania, 54% for Michigan, 53% for Wisconsin, 77% for Arizona, 75% for North Carolina, 68% for Georgia, and 61% for Nevada.

The new probability total is a nearly five-point boost for Trump since Thursday, when he was given a 58.2% probability, itself a boost from the 52.4% a week prior. The prediction shows a further eroding of Harris’s honeymoon support since President Joe Biden dropped out in July.

In a Substack post, Silver explained that the new poll was such bad news for Harris due to the large sample size and reliability of the poll, which he ranks as the second best. He said the new poll “confirms the model’s view that there’s been some sort of a shift in momentum in the race.”

The election analyst zeroed in on perceptions that Harris is too progressive as especially alarming for Democrats — the poll found that 47% of likely voters believe that Harris is too liberal, compared to 32% who see Trump as too conservative.

“I’m not a messaging-and-tactics guy like Dan Pfieffer, but I’m not quite sure how Harris is supposed to spin her way out of this perception,” he wrote. “Her convention speech was aggressively centrist and aimed at male voters, which I thought was smart. But there’s a track record here of progressive policy advocacy on the 2019 campaign trail and in her voting record in the Senate.”

Her attempt to mitigate this issue presents a problem in itself, Silver argued.

“The flip-flopping may explain why Harris has been weirdly reluctant to do media hits or articulate policy specifics,” he continued. “This strategy may have worked well enough when she was riding high off the vibes of the Democrats’ candidate swap, but it’s causing her more problems now.”

The New York Times/Sienna poll interviewed 1,695 registered voters nationwide from Sept. 3-6, with a margin of error of +/- 2.8%.