Case study of lying with manipulated numbers and measures

Wonderful example of the importance of using the right measure of success.  

This is astonishing. Almost every state that’s failing is Republican. pic.twitter.com/i8RGdEifsl

— Brianna Wu (@BriannaWu) October 12, 2020

Wu has chosen to use total Covid-19 cases per million as her measure of success.  There are plenty of alternative measures in circulation.  Covid-19 hospitalizations per million.  Covid-19 ICU cases per million.  Covid-19 deaths per million.  Excess all-causes-deaths per million.  

Why did she use Covid-19 cases per million as her measure of success?  It is among the least useful and least meaningful measures.  As testing expands, you will naturally have more Covid-19 cases per million, not necessarily because the disease is spreading but because you are looking for it harder.  

Virtually every other measure mentioned would be more revealing and useful.

In addition, there is a further tell that this is not a serious argument she is making but mere agitation.  The pandemic hit here on January 21st with our first case.  Despite that, Wu begins her data series for June 1st, 2020.  Truncated date lines almost always are an indicator that something is being hidden.  Our worst death rates were racked up before June 1 with New York, New Jersey, Massachusetts, Connecticut, and Rhode Island clocking in death rates no one else has, blessedly, yet seen again.

It is only by starting her calendar on June 1st that she is able to make the argument which she is making.  And if the plausibility of your argument is dependent on misrepresentation and deceit?  Well, of course there is going to be blow back because you are wasting everyone's time with cognitive pollution.

Rita Panahi does some education, showing her work and then even memorializing Wu's error to serve as a learning moment.

Just in case she accidentally deletes the tweet. 🧾 pic.twitter.com/oUljN4MCs6

— Rita Panahi (@RitaPanahi) October 13, 2020

And as night follows day, when empiricism meets ideological fervor on Twitter, Wu's originating tweet is (accidentally?) deleted.  

 Wu's tweet series is a good example of plausible lying by measurement selection and date truncation.  But it is also a gateway as to what the better measure would be.  Any of the others mentioned is the glib answer.

But the better answer is that we have at least four measures.  The first one we don't even discuss because we achieved our goal.  All the shutdowns were in service of avoiding a conceivable catastrophe; that the pandemic would outstrip our hospital and medical infrastructure.  We worried about that at the beginning and it was a plausible concern.  Despite the plausibility, no state suffered the overload of their healthcare infrastructure.  It is worth recalling that that was the initial measure of success and we succeeded.

There are three other measures of success though.  The second one has already been mentioned and that is the number of deaths per million.  That is pretty fundamental and so far we are better than many European countries with comparable health systems and worse than some others.  

A third measure is a measure of premature mortality or Years of Potential Life Lost.  The Spanish Flu of 1918 was so lethal and disruptive because, unusually for such pandemics, it was notable for the excess loss of life among otherwise healthy young adults, usually those least vulnerable.  Usually, with plagues, it is the very young and the elderly who are most vulnerable.

Covid-19 has also been fortunately different.  The young are virtually immune as are most people up to 50 or 60.  It is only the elderly who are especially vulnerable and even among them, death is almost completely related to those with one or more co-morbidities.  The average age of death in almost all countries has been in the seventies or early eighties.  I think I saw a report yesterday that in Britain, the average age at death for Covid-19 has so far been 82.

In some countries, such as Sweden, most Covid-19 deaths have been among the seriously ill elderly and it is notable that when you look at the yearly all-causes death figures for the past few years, they are somewhat higher this year but it follows two preceding years when all causes death were lower than usual due to mild flue seasons.  I have heard this referred to as the "dry tinder" hypothesis.

The upshot is that the years of potential life lost has been remarkably low.  All deaths are tragedies but effectively with this pandemic, given who is actually susceptible and vulnerable, in almost all cases, deaths are being brought forward a few months.  We are not losing the young and healthy.  

Finally, there is the haunting trade-off which we are still trying to understand and quantify.  Simplistically this could be straw-manned as 1) let the disease run its course, protect the most vulnerable, keep the economy functioning as close to full capacity as possible, and 2) shut everything down to try and save as many of the most vulnerable end-of-life, co-morbid endangered elderly.  In the latter instance, not only do we have deaths per million but we also have lost GDP per 1,000 deaths.

Most developed nations set a pretty high store of value on human life.  We are remarkably elastic in our willingness to spend what it takes to save people and we are extremely uncomfortable making the calculus of the economic value of a human life.

As uncomfortable as we are with that measure, it is none-the-less informative and useful in trade-off decision-making.  Just how much are we willing to sacrifice in lost national productivity to ave a life that would most likely be lost in a few months to the underlying comorbidity?  

I have not seen any such analysis but that is where I think we will end up.  Some countries/states pursued policies which took both life protection and economic sustainability into account.  Others by necessity focused on keeping the economy humming, others prioritized saving lives at all costs.

What are those rankings going to look like?  I have no firm idea but it seems at this moment that we should have followed our original pandemic policies and never shut down.  The US, overall, has probably suffered more deaths per million and a greater loss of economic welfare than we had to were we to have focused solely on aggressively protecting the vulnerable while keeping everything as open as people were comfortable doing.  

Under such a measurement system New York, New Jersey, and Connecticut are probably going to look like disasters with both high deaths per million as well as most extreme economic loss of productivity.  Massachusetts I am less certain about.  Certainly high deaths per million and probably a high loss of economic activity but not as bad as New York and New Jersey.  California will probably among the upper half of worst performers, not so much because of high deaths per million but because of self-inflicted and prolonged economic damage.  

The much derided Texas, Georgia, and Florida examples are probably going to come out pretty well n the hybrid measure of deaths per million and economic cost.  

We'll see.  We have a long ways to go and a lot yet to learn.  But learn we will, as long as we don't get distracted by the cognitive pollution of the likes of Wu.