Summary to date of my COVID-19 modeling study (Facebook comment)

I'm conducting a small study of how well the highly regarded MIT-Sloan model predicts COVID-19 deaths in 3 states, Florida, Georgia, and Louisiana. As you are aware, some people on the right argue that the prominent epidemiological models are not trustworthy because they greatly exaggerated the threat from the pandemic. This false argument is based on the fact that under the worst case scenario (that is, if no steps were taken nationally to flatten the curve), the models predictions were in the range of 200,000 plus deaths by this August.

As of me writing this comment, the predictions of all the prominent models have been adjusted down precisely because most states implemented shelter at home orders and the rest implemented at least some restrictions on their citizens. People who wish to discount the threat of COVID-19 conveniently ignore the fact that the models are adjusted continuously as things evolve, in particular depending on what steps are in place to contain the spread of infections. So, on April 16 I recorded the MIT-Sloan projections for the total number of deaths each state would have on May 1. They were: Georgia (913), Florida (1271), Louisiana (2281). Today (April 30), the official total deaths for each state (which are almost certainly undercounts according to the experts) are Georgia (1107), Florida (1268), Louisiana (1862).

Georgia has already exceeded the prediction by 21.2%. Florida is up to 99.7% of the predicted deaths on May 1 with 2 days left. Louisiana deaths are currently 81.6% of the MIT-Sloan prediction for May 1. Florida, which is taking steps to suppress the official tally of deaths from COVID-19 (see article below), will surpass the prediction by May 1. Only Louisiana will have fewer total deaths by May 1 than the model predicted 2 weeks ago. Why? Because Louisiana has implemented the strictest restrictions and is keeping most of them in place. Georgia and Florida have relaxed their restrictions and consequently more people are dying there. If you look at the current MIT-Sloan predictions for these 3 states moving forward to August, you'll see they have been adjusted up for Georgia and Florida and down for Louisiana.

I conclude here than this model did not exaggerate its predictions for these 3 states. It underestimated deaths in Georgia and Florida. The moral? Be like Louisiana. Don't be like Georgia and Florida.