COVID-19 Exponential Growth forecasting Induction It's a big building with patients but that's not important right now Life nonlinear dynamics real world Risk

Apples to Big Apples

It is one of those sad aspects of politics that people become so attached to that they loose sight of reality. The POTUS and policy makers who are asking us to stop social distancing are up against a lot of data that refutes their suggestions; with the POTUS now listening to the inner circle of sycophants who know a good opportunity to get their way when they see one, we get late night ALL CAPS tweets from Tweety-Amin.

It is not the case that there isn’t room for nuance in our social distancing metrics. It is not the case that we cannot come up with, over time, a better risk stratification methodology to get America back to work. It is ABSOLUTELY the case that we do not have enough data to engage in the analysis of lifting/changing the current policies.

Put another way, the general (non-naive) precautionary principle delineates conditions where actions must be taken to reduce risk of ruin, and in these cases, the traditional cost-benefit analyses must not be used. A lot of smart people have gotten into the national argument on Medium and Twitter. They are making their cases well. However, they are using traditional cost-benefit analysis to support their positions.

There is already, and there will be more, harm done to our nation’s economy. Worse for the world, as we have all just learned, the robustness of our respective economies is not as strong as we had been pretending under globalization. Our economic models were missing something central: the risk of a big epidemic (pandemic). As we now see, such a risk can quickly outweigh the gains realized from distributed supply chains and cheap labor.

At this point in time, we don’t know much about how COVID-19 is impacting different populations here in the United States. If we approach this from a naive emperical point of view, we can use the data we do have to disprove ideas and theories, but not prove them. We can use the historical evidence we have from other countries’ COVID-19 experiences to refute conjectures, but not to affirm them.

When it comes to public policy, we will likely do best by dividing our theories about how best to approach COVID-19 into two camps:

  1. Theories that are known to be wrong, as they were tested and adequately rejected.
  2. Theories that have not yet been known to be wrong, not falsified yet, but are exposed to be proven wrong.

I think we can easily reject the notion that 14 days of social distancing is enough to catch all the people that are infected with COVID-19 and whom might spread it. In the first place, there is some portion of the population that is infected and is asymptomatic (no outward symptoms, Typhoid Mary anyone?). Then we have the issue that from the data in China we see a long tail for onset of first symptoms. As we learned from that study, as outlined in the graphic below, there is a small population who will not experience symptoms until after 14 days.

So, conjecture of when might be a good time to encourage a relaxation of social distancing we have at least two data points that refute the notion that 14 days is the correct time to relax the policy. Trump’s proposal that 14 days is enough therefore should be rejected.

There might be a different number of days when relaxation makes sense. Someone might propose a different number of days. Then, we can collect the data needed to falsify the new number. We can analyze our results with the full understanding that the best we will ever be able to say is that the theory is not false at the moment.

Today (March 24, 2020) Trump is talking about “Opening everything up by Easter” – which means NY will have been social distancing for a total of 22 days. This is certainly a desirable goal from the perspective of people at low risk, and those just sick and tired of being cooped up. However what do we think that is likely to mean?

We can look at what happened in Hubei (China) (an environment where very strict controls were put into place), to see what their situation looked like 22 days in. Spoiler alert, it took 45 days for Hubei to get their R naught (rate of infection growth) down to <= 1.0 (at 115/100,000). Models are indeed just models. With that in mind, the linear model for Hubei, using the data from the first 22 days, does intercept the peak measured rate for Hubei within 2 days of the 45th day.

Two days later, things aren’t looking much better …

None of this is to say that we have any idea where New York State will peak based on these models. Models rarely (if ever) have all the inputs that would be required to make a completely accurate inductive conclusion. Knowing that it took Hubei 45 days to achieve an R naught of <=1.0, it seems highly unlikely that the smart policy for the US is to decrease social distancing by Easter.

Being human, it is not unusual for us to want a defined set of parameters to help us understand where we are in a situation. I’m sure Trump is exceptionally uncomfortable with the truth of the matter, specifically that we do not know when it will be appropriate to modify our social distancing policies.

With the International Labour Organization (ILO) projecting that the effects of the COVID-19 pandemic could destroy up to 25 million jobs around the world, one has to wonder about Trump’s motivations. I mean not really, I think most people know, but perhaps some of his supporters might get “woke” if they stop and think about it.

We all know that Trump owns seven U.S. hotels, including three — in Washington, Miami, and Chicago — with outstanding loans. The original value of these loans was more than $300 million. So, with the POTUS not having given up his business interests, we can only assume one of the inputs to his decision making is how much the Trump Organizaton stands to loose.

Again, traditional cost-benefit analysis (as done by business people and economists) is not the right way to go here. Making false comparisons of apples to oranges is not helpful either. Yet, on so many of the Trumpian information networks there is an attempt to do just that.

Trump and others are using car accident rates in an attempt to confuse the issue using analogies that are simply not comparable:

You look at automobile accidents, which are far greater than any numbers we’re talking about. That doesn’t mean we’re going to tell everybody no more driving of cars. So we have to do things to get our country open.

Donald J. Trump

Nothing shows how mathematically impaired our stable genius president is than assertions like that. SARS-CoV-2 is not a car that gets into an accident. It is a virus and infects and kills.

To get to a proper Apples to Apples comparison, we can do a little thought experiment. One that will illustrate a fair car accident vs COVID-19 scenario for policy makers to consider. A simulation where we control the inputs.

Scenario: Let’s imagine a future world of 100% self driving cars. We now have a software virus which has infected the self driving cars and through the regular information exchange between vehicles it has found a way to replicate. It is designed to take over driving and deliberately cause accidents after residing dormant in the car’s computer memory for between 5 and 21 days. Some of the accidents cause injuries. We are now in possession of a test which can determine with great accuracy which cars are infected but, due to privacy and security concerns, a technician must come to your house to test your car. It will take time to figure out where all the currently infected cars are.

To simulate the rate of change in accidents, we’ll start with 100 accidents on the first day (the typical amount of accidents in the US on any day that cause injuries) and use the increase in confirmed COVID-19 infections from NY state to model the rate of infection of cars with the injury accident causing software virus. After we no longer have real data, we will then use the average the daily increase to project out into the future. Here is the result starting with the original 100 injury accidents of software virus infected cars:

As you can see, simply with the power of exponential growth, by Day 28 we will have had more injury accidents caused by this software virus than we do ALL car accidents in a typical driving year (more than 3 million).

So the question for policy makers is this: on what day do you decide to enforce social distancing of self driving cars by making these cars stay off the streets? By day 5 you had a 300% increase. The TV news stations would be going crazy. At that point, I would think it was a no brainer.

Clearly, policy makers would look at the exponential growth and act quickly. Further, we probably would not allow our self driving cars back on to the road until we could confirm each of them was software virus free. This is much more of an Apples to Apples comparison than what the POTUS and his FauxNews cronies are offering up.

In conclusion: (a) data can only be used to refute assertions not prove them; (b) the POTUS and other policy influencers/makers who are asking us to stop social distancing are up against a lot of data that refutes their suggestions; (c) continue to stay the fuck home until we see an R naught of less than or equal to 1.

I’ll try to write another one of these on stratified risk strategies later this week.

Leave a Reply