US Election 2024: Poll aggregator review - Silver Bulletin

Old dog, new tricks

The Silver Bulletin model is one of two clear descendants from the venerable FiveThirtyEight model, which shot to fame by predicting the results the 2008 election with remarkable accuracy … and then became notorious after the 2016 election, despite the fact it was righter than almost anyone else.

In 2016, it’s worth emphasising one more time that FiveThirtyEight - at that time run by Nate Silver - was almost uniquely cautious in predicting a Clinton victory. FiveThirtyEight gave Trump nearly 30% chance of victory, while other forecasters had him down in single figures, or even negligible chances. Nonetheless, they were hit by the brand-leader curse: they were the most famous, and so got a lot of heat for the collective failure.

In fact, due to some IP (intellectual property) implications of Nate Silver being forced out from Disney/ABC’s ownership of the brand, it’s plausible to hold that this new Silver Bulletin model is the old FiveThirtyEight model (with some tweaks), and the replacement model running under the FiveThirtyEight brand is actually more different - closer to a new challenger.

So, while others have got some intermediate elections “righter” than this model, particularly at the state level, let’s provisionally acknowledge that the Silver Bulletin model, or at least its methodology, has a good-to-excellent track record, which has held up fairly well even when the polling errors became particularly challenging. And that, given this, it is very worth understanding what the model has to say this time around.

---

Outlier territory

And the short answer - as of today: 30th August 2024, is that the Silver Bulletin is the most bullish of all on Trump - giving him a 55% chance of victory, while the new version of FiveThirtyEight is at the opposite end, giving him only 43% - a spread of >12 ppts between them.

The odd thing about this is that both these models operate a broadly similar methodology, on basically the same data. It’s a very solid, battle tested methodolofy, and despite the differences in the results, Silver and (new) FiveThirtyEight are far more similar to one another than to some of the others that fall between them in this table.

It’s worth understanding how they can come out with such different results.