I’ve told you a few times about Professor Michael McDonald’s early vote analysis. He has a paywalled final analysis of the early vote in North Carolina. The upshot is that by conventional early vote analysis, Donald Trump appears poised to win North Carolina. That wouldn’t be a surprising result either on the basis of history or the current polls, which show a dead heat race with the slightest advantage to Trump.

But McDonald also notes that it is an unusual cycle with conflicting signals. The polls look more favorable to Harris than the numbers in the early vote. Actual votes matter more than polls of votes, by definition. But this is a reminder of what early vote analysis is based on. We’re largely going on party registration and limited demographic markers as a proxy for voter intention. Those will generally point in the right direction, except when they don’t.

McDonald notes the paradox that while there are more early Republican votes than usual, which should compress the gender numbers, the gender spread has remained quite high — 10.5 percentage points. The story from the weekend’s Selzer poll was that Republican women, particularly older women, and unaffiliated women were breaking hard for Harris at the end of the campaign. If that trend is real and operating outside of Iowa you can plug that assumption into these early North Carolina numbers and paint a rosier picture for Harris. McDonald concludes by saying he’ll stick to conventional prediction while noting that there’s a real argument for the other scenario.

If you’ve followed my thinking on this you know I’ve long had a pretty low opinion of political betting markets. Their user base tends to lean right, with the built-in bias you would expect that to cause. They’re also prone to manipulation. But the biggest problem is that, in my view, they’re largely derivative of polls and the press narratives. Garbage in, garbage out. I will simply note that the wild gyrations all of them have been doing over the last three or four days provide, I think, some backing for my argument.