Why Coronavirus Models Fail

It’s easy to be seduced by the word, “model.” Alas, we have short memories. Just weeks ago, respected models predicted cataclysmic shortages of ventilators. That spurred the media to attack government for not doing more. Yet just a week later, the creators of the model policymakers most rely on and that the media cites, the IHME model, essentially said, “Never mind.” The model’s prediction of a lack of ventilators and hospital beds proved dramatically wrong.
The data is too subject to invalidity: too few subjects, not a random selection of subjects, assumptions about the percentage of people who will follow social distancing rules, etc. Most eviscerating to models’ predictive validity is what I call the monkey wrenches: the factors that emerged between the weeks when the data was collected and when policymakers act on the model’s recommendations.
Lest you think the monkey wrenches are few or minor, here are those I learned about just this morning:

NPR reported that most people on ventilators die and most of the survivors are very old and unlikely to have much quality of life for much longer.In a separate report, NPR reported that the coronavirus antibody tests produce many inaccurate results. That means that carriers can march around oblivious to infecting people and that people not-infected are worrying needlessly.CNN reported that a team of Harvard scientists concluded that unless a solid vaccine is developed soon, we’ll need sustained or at least intermittent social distancing not just for months but until 2022. That increases the likelihood of increased problems from domestic violence to supply chain shortages, which could cause the benignly termed “social unrest,” more clearly described as looting and home invasions.Speaking of supply chains, The Institute for Supply Management reported that in just the last two weeks of March, the most recent period studied, supply chains of input manufacturers doubled.CNN reported today that 2,156 New York City cops have tested positive. That’s up from just 100 as of March 24. As of last week, 20% of all New York City police officers were out sick.Politicians and the media have been reporting a flattening of the curve. But today, whoops, he highest daily number of US corona deaths ever.Health care workers, the group most likely to be wearing PPEs, are dying at surprisingly high rates, higher than government-reported data indicates.

Multiply what I’ve learned just this morning by the rapidly unfolding situation, and we must very carefully weigh the economic costs of major expenditures and incursions on human daily life against what the models, created a week(s) earlier, says we should do.
Especially in macro decision-making, logic, subjective experience, and yes, data, must all get serious consideration in policy-making.
My personal integration of data, logic, and observing qualitative experience suggests that regarding the coronavirus, we should:Reallocate government resources to support vaccine development and at-home-self-testing (like a pregnancy test) in small, promising companies that are undercapitalized. Focus education and policing efforts regarding social distancing by census tract. Only a tiny percentage of tracts have low compliance.Jurisdictions with low-incidence and a flat curve, especially those in which there are many jobs, should be reopened, but with social-distancing measured urged or required. Large venues such as nightclubs and sports arenas should be closed until a solid vaccine has widely been implemented.Instead of printing more money and increasing existential debt, return to the taxpayer all the money that is being spent on corona handouts, which are heavily fraud-ridden: e.g., self-reported need for taxpayer money.