Update: @LizSpecht does some similar math projecting the implications on the healthcare system.

I like to think I am pretty good at math. Day to day stuff like calculating tips in my head, but also more obscure things like calculating surface areas of 3D objects, doing basic calculus and developing basic math models to fit data. I’m still in awe when someone on the academic team I coach evaluates a quadratic integral with a log thrown in in less than 5 seconds, so I am by no means extraordinary, but pretty good. And yet, I am absolutely terrible at statistics. I have gone as far to say that since I am pretty good at math and bad at statistics, that statistics must not actually be math, which is no doubt some taxonomical form of the fundamental attribution error.

I suppose I can take some solace in the fact I am in good company – or at least there are a lot of other people equally bad at statistics. I would argue one the main points of Thinking Fast and Slow (still one of the best books I have read in the past 5-10 years) is that many of the problems we face, both personally and communally, come from the shortcuts we create because we are bad at thinking statistically.

Remembering that I am bad at statistics, it seems to me that we are all suffering from a bit of being bad at statistics in some of the discussions that are going around about COVID-19 / Corona Virus. I’ve seen a few posts from credible sources that are trying to downplay the potential impact of a COVID-19 pandemic by stating how many people have died from the seasonal flu vs how few (comparatively) have died from COVID-19. I’ve even seen a few (from perhaps less credible sources) that are making comparisons to the number of people that have died from cancer, diabetes and even car crashes.

In one way, these sorts of arguments are doing some good in that they are fighting against one of the heuristics that Thinking Fast and Slow mentions: since we are such bad statistical thinkers, we substitute the ease that we can recall something from memory as a proxy for how likely it is. With all of the news coverage we are likely all overestimating the chances of dying from a raging case of Corona.

What these arguments miss is the population size. While its true that the total number of people that have died from seasonal flu is much greater than the number that have died from Corona, it’s equally true that the number of people that are infected with seasonal flu is also much higher…for now. What matters here is the rate, which a real statistician would argue is barely statistics, but we still seem to be getting wrong.

There are two rates that matter: The infection / transmission rate and the mortality rate. The latter seems to be hovering between 1% and as high as 3.4% depending on which model you use, which is between 10 and 20 times more deadly that seasonal flu. The former is what seems to be the big question. Infection rate depends on things like virulence, how quickly and how often symptoms show and actually mortality (if it kills you more often / quickly, it is less likely you will transmit it to others). If we have an infection rate that is 10 to 20 times less than seasonal flu, then statistically we will get the same number of deaths as seasonal flu. Unfortunately that doesn’t seem to be the case so far with some officials stating that between 40 to as much as 70% of the worlds population ultimately destined for infection. If the low end of both numbers are right that’s 30.8M people dead by the time this is all over. Let’s really hope I am as bad at statistics as I think.