Governments are (very likely) irrelevant

 

Although “action bias” and “political partisanship” all but guarantee that people will be making governments responsible for the consequences of the pandemic at a national level, I do think that drawing conclusions on government policies from the number of deaths (or any other relevant covid19 related variable) is a very risky business:

1.- Deaths per million is a variable with a "probability distribution"

2.- The shape of this distribution very likely depends on several drivers that can be clustered into two main categories:

• "Non-actionable drivers": like average age, population density, genetics, prevalence of pre-existing conditions, "base" level of people interaction (very likely higher in Puerto Rico than in Stockholm), etc.
• "Actionable drivers": like lockdowns, mask wearing, rapid test availability, social distancing, health system preparedness, etc...

3.- These drivers "define" the shape of the distribution, but the distribution, in any case, follows a "power law" pattern, which means you can have very extreme differences in "observed values of the variable" even within the same "distribution" (meaning "for the same set of policies").

If this is the case:

a) The real shape of the distribution(s) is very difficult (or impossible) to know from the available sampling(s)

b) Even if the distribution could be derived, the "average" value is meaningless to draw any conclusions with practical implications. The "political effort" to "manage the distribution" (and the criticism or praise of this effort) must focus on limiting the risk of "very fat tails", not in managing the "average" or higher order statistics. 

c) Even for the optimal “politically managed distribution" very high "observed values" of the variable can happen and the overlapping with the "very poorly managed distribution" observed values could be very significant.

d) This makes very difficult to form robust judgments on the real effectiveness of the measures adopted by the different governments, since they affect the "shape" of the distribution which is almost impossible to observe.

e) The policies adopted by the government, only marginally affect the "actionable drivers", since, even in the absence of "mandatory" lockdowns or facial coverings, people would be taking measures and this “voluntary” measures will be self-balancing: they would be more radical the more deadly the outbreak.

If this is the case, the government policies would only marginally affect the "actionable drivers" which are themselves only a subset of the drivers (difficult to know how relevant compare with the non-actionable. My guess would be "less relevant") and, still, even for the same "shape of the distribution", very significant differences in the "observed values" could happen.

Government policies, as always, are much less relevant than voters tend to think and their actual consequences (intended or otherwise) extremely difficult to stablish. 

Political accountability is an illusion and will be (once again) very easily avoided by the cunning politicians. As Tetlock points out, we will be, once again, surprised by "the ingenuity and determination that political elites display in rendering their positions impregnable to evidence".

*************************************************************************

A couple of articles on this line:

https://www.santafe.edu/news-center/news/transmission-t-024-cristopher-moore-on-the-heavy-tail-of-outbreaks

https://nassimtaleb.org/2020/05/tail-risk-contagious-diseases/