Kamalamentum vs. economic headwinds

So far, the vice president is still coming out ahead.

In general, the Silver Bulletin forecast has tracked prediction markets extremely closely — but there’s more of a divergence now. Harris rose again in our forecast today: in the span of a week, she’s gone from a 37 percent chance of winning the Electoral College to a 53 percent chance. The race is still a toss-up for all intents and purposes, but if you’re sweating the details, you’d rather have the 53 percent side of a bet than the 47 percent half.

At Polymarket, however, the odds are roughly the reverse of this, with Trump at 54 percent and Harris at 44 (and 2 percent for Michelle Obama for some reason). That may be because of economic news: Friday brought a disappointing jobs report, and the S&P 500 was off by 3 percent on Monday, so pour yourself a glass of something strong before you check your 401K. Prediction market traders have heavy adjacency to financial markets, so perhaps it’s not surprising that they’re sensitive to this stuff.

But are they right? Is the model just a bit out over its skis on Harris? Or would you have a profitable bet on her, if you were into that kind of thing?

What I’d say in general, before we go behind the paywall, is that I think this is a somewhat challenging time from a forecasting standpoint. There’s been a lot of huge political news that’s piled on top of itself. I’m not sure that the polls have yet entered some sort of steady state — instead, Harris has fairly consistently been rising, and is now ahead by almost 2 points in our national polling average.

Usually, I’m not a big believer in “momentum” in polling. The model is supposed to be carefully designed to react by the “right” amount to new polling – that is, neither overreacting or under-reacting. (A more technical explanation: if they’re calibrated properly, then the polling averages shouldn’t be autocorrelated from day to day, which means that they efficiently incorporate all information and the best prediction of tomorrow’s polling average is today’s polling average.)

But it’s possible that this assumption is violated by Harris’s late entry into the race. After all, the sample size of something like this happening is zero in the data that the model was trained on.