Election Forecasts, Schmelection Forecasts

What if pollsters are just too afraid of Trump this time around?

Out of curiosity about the data-science side of things, in September, I started my own little forecast for the 2024 U.S. presidential election, which considers three different scenarios: ① based on weighted high-quality polls only; ② adjusted by the average polling bias from 2016/20; and ③ adjusted by just half the 2016/20 bias. I’ve previously written about my methodology and predictions here and here.

In the latter article, I lay out why I believe another polling bias underestimating Trump is a very real — even probable — possibility that can’t be dismissed. Trump as a candidate seems to completely defy the laws of “normal” elections. It’s unclear how much pollsters have understood and fixed the reasons for their misses in the past 2 elections; the “shy Trump voter” might still exist. On top, Republican-initiated voter suppression — like purging voters from voter rolls in Virginia this close to an election, as recently allowed by the Supreme Court — will disadvantage Democrats in at least some places, which can’t be captured by opinion polling.

In the past weeks, however, I’ve also seen arguments as to why things could be the other way ‘round this time, and I must admit that I find some of them quite convincing.