Reasonable bounds on presidential election outcomes

How can we know how wide the tails should be?

FiveThirtyEight launched their 2020 election model this week. It’s much less certain than the alternative available, such as that from The Economist or a lesser-known forecaster named Jack Kersting. 538 gives Biden a 72% chance of winning, and the others are in the mid-to-high 80s.

Nate Silver has offered up a ton of justification for why his model is this uncertain about the election. I don’t want to get into who is right or wrong right now—it’s very hard for me to do that objectively since I have a pretty strong incentive to disagree with Nate—but I do want to talk briefly about how we can calibrate our expectations for possible electoral outcomes based on other sources of data, and then gauge which forecasts are more likely because of that.

To introduce the topic playfully: One of the things that I think is funny about the 538 forecast is that they have this little cartoon character that treats you like an idiot in order to explain the model to you. His name is Fivey Fox. I do kind …