As a very rough estimate, suppose that we value life at $10M and that long covid makes life 10% worse. Then getting long covid is like losing $1M. And in getting covid, there is a 1⁄500 chance of losing that $1M (amongst other things), so an expectation of losing $2k.
So what does this mean for everyday activities like indoor dining? Indoor dining costs something like 5k microcovids, or a 1⁄200 chance of getting covid. Getting covid is bad for reasons other than the possibility of getting long covid, but if we just focus on that long covid component, then it’s a 1⁄200 chance of losing $2k, or an expectation of $10.
On the microCovid math: I noticed right away that number (1/200) didn’t make sense, so I did some quick math. Average American used to eat out 5 times/week, and most people are mostly back to normal. So this would be saying something like 1%-2% of people per week get Covid right now from indoor dining, which doesn’t make any sense, that’s more than double the number of verified cases from all sources. Something screwy there.
Huh, good point. The default assumption for indoor dining on microCOVID is that there are 15 people within 15 feet of you. That seems a little high to me, but going from 15 to 10 only cuts the risk by 50% or so. Not enough to explain it. Even the low end of their confidence interval is around 2k microcovids.
Many americans have (recently) had covid, making them (more or less) immune. That would cut expected microcovids for a random american by some notable factor.
Wait, really? This is a shocking number to me—I assume you got the figure from here or here (I assume both refer to the same study, but neither link to it).
More than 56% of those who eat out will do so at least 2-3 times per week—including in-restaurant dining and ordering food to go.
That is, conditional on being a person who eats out, P(>2-3 times/week) >= 56%. This seems difficult to square with either mean or median person eating out 5 times per week while the 90th percentile is 4-6 times per week.
Yeah I was working super quickly, and Google was being remarkably useless in giving me the actual number of meals so that was the one concrete answer I found. Curious if anyone else has better fu.
I agree 5 sounds like a lot but it’s out of 21 meals so it seems very not-crazy to me. Even if it’s 2, numbers still don’t add up—we don’t see indoor dining restrictions on their own cutting transmission rates by gigantic amounts.
Long Covid often leads to an inability to do substantial amounts of deep work. Given that a lot of people in this community derive a big part of their self worth from being able to do specific kinds of concentrated work (whether that’s healthy or not) and work sooner counts more than work later,* I think 10% is maybe a bit low. (It depends how quickly you expect Long Covid to improve. If it’s 70% impairment for 2 years, would that translate to “10% of remaining life,” impact-wise?)
*Due to discounting for haste consideration, cognitive decline from eventually getting rigid in one’s thinking, early AI timelines in some people’s cases.
As a very rough estimate, suppose that we value life at $10M and that long covid makes life 10% worse. Then getting long covid is like losing $1M. And in getting covid, there is a 1⁄500 chance of losing that $1M (amongst other things), so an expectation of losing $2k.
So what does this mean for everyday activities like indoor dining? Indoor dining costs something like 5k microcovids, or a 1⁄200 chance of getting covid. Getting covid is bad for reasons other than the possibility of getting long covid, but if we just focus on that long covid component, then it’s a 1⁄200 chance of losing $2k, or an expectation of $10.
On the microCovid math: I noticed right away that number (1/200) didn’t make sense, so I did some quick math. Average American used to eat out 5 times/week, and most people are mostly back to normal. So this would be saying something like 1%-2% of people per week get Covid right now from indoor dining, which doesn’t make any sense, that’s more than double the number of verified cases from all sources. Something screwy there.
Huh, good point. The default assumption for indoor dining on microCOVID is that there are 15 people within 15 feet of you. That seems a little high to me, but going from 15 to 10 only cuts the risk by 50% or so. Not enough to explain it. Even the low end of their confidence interval is around 2k microcovids.
Many americans have (recently) had covid, making them (more or less) immune. That would cut expected microcovids for a random american by some notable factor.
Wait, really? This is a shocking number to me—I assume you got the figure from here or here (I assume both refer to the same study, but neither link to it).
However, this article has data that strongly suggests the real number is much lower (based on this underlying survey).
Core claim:
That is, conditional on being a person who eats out, P(>2-3 times/week) >= 56%. This seems difficult to square with either mean or median person eating out 5 times per week while the 90th percentile is 4-6 times per week.
Yeah I was working super quickly, and Google was being remarkably useless in giving me the actual number of meals so that was the one concrete answer I found. Curious if anyone else has better fu.
I agree 5 sounds like a lot but it’s out of 21 meals so it seems very not-crazy to me. Even if it’s 2, numbers still don’t add up—we don’t see indoor dining restrictions on their own cutting transmission rates by gigantic amounts.
Microcovids start to diverge from actual probabilities earlier than you might think. I’ve commented on their github page
Can you link to your comment?
Long Covid often leads to an inability to do substantial amounts of deep work. Given that a lot of people in this community derive a big part of their self worth from being able to do specific kinds of concentrated work (whether that’s healthy or not) and work sooner counts more than work later,* I think 10% is maybe a bit low. (It depends how quickly you expect Long Covid to improve. If it’s 70% impairment for 2 years, would that translate to “10% of remaining life,” impact-wise?)
*Due to discounting for haste consideration, cognitive decline from eventually getting rigid in one’s thinking, early AI timelines in some people’s cases.