>RTLD: 41 cited deaths among 3M grocery workers is a lower death rate than the population as a whole ?

If you're comparing grocers death rate against the population as a whole, that's not a good comparison as it doesn't control for grocer's demographics. I think that grocers would be younger (on average) than the the general population, and therefore should have lower death rates.

What is the term for this sort of statistical error? I see it very often in online discussions.

I think it's related to Simpson's paradox [0] or general issues with interpreting conditional probabilities, but I'm not sure what the best overarching term is for this sort of fallacy.

[0] https://en.wikipedia.org/wiki/Simpson's_paradox

Failure to control for confounders.

Grocery store workers aren't going to be looking at aggregate statistics to make their decisions or decide how they feel. They are going to be basing it on "Joe in produce died and we think it was Coronavirus."

Then the entire store will be scared and it's not unreasonable because they probably interacted with Joe in produce at some point recently.

A lot of people aren't getting confirmation by testing. Most people will self diagnose based on symptoms. This means that official data undercounts deaths from the condition and unofficial data overcounts death from the condition, but this is a case of "which way do you bet?" And when faced with a pandemic, the way to bet is "That guy with that cough who is now dead probably died from the pandemic. No, I don't need official testing to be scared that I'm next because I pulled a shift with him recently."

> because they probably interacted with Joe in produce at some point recently

> because I pulled a shift with him recently

It's more likely they are scared of their customers rather than Joe. Because Joe is just a person who contacted with contagious clients and they will never contact him again. They've probably already sanitized their store. But they will have to contact other clients who infected Joe and there is nothing they can do to avoid that.

That may be true, but people tend to assess risk in a not entirely rational way.

https://www.psychologytoday.com/us/articles/200801/10-ways-w...

> Have I made a mistake somewhere?

"At least" 41 deaths among 3 million grocery workers.

I can't imagine we have anything resembling complete data for the number of grocery workers who have died. "At least 41" could be 42; it could be 500. Trying to draw conclusions on the relative safety of a grocery store employee versus the general public from that number is crazy.

> Trying to draw conclusions on the relative safety of a grocery store employee versus the general public from that number is crazy.

That's exactly what the Washington Post article is doing.

Is it 14 per million in addition to the normal 22 per million? Then it would at least show up on the top ten causes of death.

Kevin notes some more weird things about the coronavirus here: https://www.motherjones.com/kevin-drum/2020/04/something-abo...

I'm immediately curious about the actual age distribution. Drum compares this to mortality for people between 20-65. My guesstimate is that grocery store workers are on average younger than the general population. I also wonder how many of the older grocery workers are taking time off.

3M protects from the coronavirus