Thread by @Doc_0

We'll be arguing about these coronavirus forecast models for years to come, and it's clearly important to refine them as much as possible so they are more accurate next time. It's strange that anyone would still be trying to argue the models were right this time. They weren't.

Why the models were wrong, and by how much, should be a topic of serious discussion, but at this point insisting on their accuracy is utterly quixotic. If a model that accounted for full social distancing said cases would be 4x or 6x worse than they actually were, it was wrong.

It doesn't matter if the model was "adjusted" as better data came in. That's fine and important to do, but it doesn't change the fact that the original model was wrong, and it's beyond weird to hear people trying to make that argument.

Weighty decisions were made based on the ORIGINAL projections. It doesn't matter that version 5.3 of the model finally begins tracking with reality four weeks into the crisis. That might mean the models will work better during the next epidemic, but that's not much comfort now.

If the models were way off, even with full social distancing factored in, there are three logical reasons why:

1. The core assumptions of the model were flawed

2. Bad data from this pandemic was fed into the model

3. Social distancing works a LOT better than the model assumed

There is also the possibility that the projections were skewed by human bias or even deliberate malfeasance, but we should expect a high level of evidence before reaching that conclusion. If we're focused on improving the models, we should carefully evaluate their performance.

We use modeling to make many decisions with huge financial costs and heavy ramifications for individual freedom. It matters very much if they're wrong, and we should be utterly relentless in establishing why and improving them. No inaccurate model should have a single "defender."

If your map software tricks you into driving off a cliff, you won't be in a mood to hear assurances that the model reasonably assumed there was a bridge, and the next update will refine the model to show that the bridge is out. None of that makes the original map any less wrong.

One thing to consider is that we don't know for certain how accurate the models of the "road not taken" were. They were wrong about how things would look with social distancing. That doesn't NECESSARILY mean they were totally wrong about how things would look without it.

Given the immense cost associated with lockdowns, it is perfectly reasonable for us to want more accurate models so we can evaluate whether the most severe actions are necessary, especially when dealing with the entire country rather than individual hot spots.

The cost of lockdowns absolutely does include lives lost, and that human damage will keep accumulating for some time after the lockdowns are lifted. It's not a binary choice between "saving lives" and "making money."

breitbart.com/national-secur…

Serous crises rarely present us with good choices, just a range of bad ones. There are costs associated with everything we could do. We need those costs evaluated with the highest degree of accuracy possible so we know what we're getting, and what we're sacrificing.

There is a tendency to excuse excessively dire predictions because hearts were in the right place and under-reacting to a severe crisis is worse than over-reacting, especially from a political perspective. The aftermath of the Wuhan virus will demonstrate how wrong that is. /end