A few words on prediction models....
Prediction models, whether they are economic, corporate, climate change, or coronavirus, are not meant to be predictive, much less accurate.
Prediction models are useful, but they're not accurate.
Probably 95% of the public, and clearly 100% of the media (200% for Bill Hemmer, so feel free to laugh out loud at him when he talks about the models), thinks these corona models are supposed to be accurate predictions of the future. Think about that - predict the future. You are already starting to hear it as the numbers come in and are beginning to settle... "The models are wrong!" (No crap, Craplock, they weren't meant to be right). And there will be a few numbnuts who go, "Well of COURSE the models are wrong! They predicted millions of deaths so that Trump would look good when they ended up being lower!" OK, with those people, you can just dismiss them with a sold 2x4 up the side of the head, because they're too stupid to be walking around without a chaperone.
Prediction models are intended to illustrate possible risk, but more to the point they are sales tools, persuasion techniques, to get people to change their behavior. That's it. That's all they are. Useful, but not accurate.
If the experts had predicted right up front that the virus would kill 50,000 people, and they told us that, we'd think, "That's not so bad. Car accidents do that, the regular flu does that, drug overdoses do that." But the experts also know that if the American people think it's no big deal, no steps to mitigate it will take place, so the experts need a little persuasion. They need a picture. Then need a prediction model that looks as scary as it can get, because if it's accurate, then the experts are gonna get blamed for every single death above 50,000. And if it's scary, they can convince the leaders to create policies to mitigate the virus, and the policies will rain down upon the people and the people will act.
The secret of prediction models is, when making one, you always start with the answer. It's either the answer you want, the answer your boss wants, or the answer the experts want. It's not like you plug in known data, assumptions, variables known and unknown, and then out the other end you get what you get. Oh, no, no, no. You start off with the answer, and then plug stuff in, and then tweak it until it gives you the desired answer.
Oh, sure, you can predict the future where things are known. Like a vehicle repair estimate, or home construction, where things are almost exactly like the past. But you can't predict the future where the variables are unknown.
So to recap, climate change models, corporate forecasting models, economic models, Coronavirus Scoreboard of Death models, they are all designed, explicitly, to alter behaviour, not to be accurate. If people could make accurate prediction models, they wouldn't be making them, they'd be sitting on their trillion dollar yachts, earning 200 percent.