I know the Washington post opinion column isn’t the right place to post numbers, but do you have ballpark estimates for how costly (economically and/or in terms of human toll) lockdowns will be in low income and middle income countries?
I do think that some people (not saying you are one!) often underestimate the human harm of getting covid-19 in developing countries (eg, they’ll quote widely discredited numbers for IFR like .1%, which obviously is ~impossible).
So it’d be helpful to do ballpark Fermi estimates for the cost of different interventions (or not doing those interventions) vs the benefits, either for the world as a whole or a specific country in mind.
I can possibly help provide the modeling on the covid side, but I don’t have a good grasp of the “cost” side of lockdowns at the moment.
Sorry for the slow reply! I had been working on some rough estimates for total (i.e. including medium- and long-run downstream impacts) costs and benefits of e.g. lockdown vs targeted social distancing, but even in high-income countries this is hard! This paper from Layard et al (using well-being adjusted life years) is perhaps the closest I’ve seen:
Happy to consider collaborating on something for developing countries, if only to get a sense for which dimensions are likely to be first order and hence worthy of further study, but I’m hesitant to believe even that would be feasible with confidence. Perhaps important enough to try in any case? Also not sure I am best placed for it, as a micro economist focusing on individual behavior...
Curious to hear what you think is the best existing evidence for IFR. Indeed 0.1% seems too low overall, but for under-60s my sense was that it is probably 0.1-0.2%, as in these papers:
Of course the quality of health system resources will affect the IFR, but neither of the ones above (China, Italy) are from ideal situations either so I honestly don’t know how much worse it will be globally.
I think ~1.1% (with fairly wide uncertainty) is a fairly realistic guess for a global IFR (including all age ranges). I basically don’t buy that the balance of factors would necessarily favor poorer and younger countries over richer/healthier/older ones, though it certainly is possible.
Here’s a preliminary document listing why I believe this. Usual caveats of being a non-professional apply, and also the tone is a bit sharper than I’d use on the EA Forum (basically the intended audience was other amateur forecasters so there are certain stylistic differences, especially around caveats).
~0.1%, or even slightly lower, seems believable for <60s in some rich countries but I don’t think you want to extrapolate age-structure arguments too strongly to novel situations (in essence I think age is a biased estimator whereas something like crude death rate may not be), and if you want to look at specific countries you’d want to look at a bunch of known comorbidities*; eg per capita, Nigerians die of heart disease at ~1/10 the rate of Indians.
One thing that I didn’t mention in my document above is that even if .1%-.2% is a realistic IFR for young people in developing countries, and developing countries are skewed young, the full IFR in developing countries will likely still be much higher.
For example, Guayas province in Ecuador has had ~11,561 excess deaths from the beginning of March to mid-April (base rate is ~3000 in that time period). My understanding is that close to all of it is directly due to covid-19 (I talked to people from Ecuador and if there was mass starvations or a different epidemic that accounted for even 2x all-cause mortality I’d have heard by now). The population of Guayas is ~3 million, so this is already a lower bound of ~0.39% of the entire population(!), and I really don’t buy that anywhere near 100% of Guayas were infected as of mid-April (or more accurately late March to account for lag between infection and death).
Ecuador has a median age of 27.9, a life expectancy of 76.6, and a GDP per capita of $6400, so definitely not unusually old or unhealthy by middle-income country standards.
This is great—thanks. My belief certainly wouldn’t be that simply because of the age structure IFR is going to be lower in developing countries. I do think that will be protective, but I also think that poor health systems will (obviously) go the other way. Some risk factors (generally more stressed immune systems) may work against them; other risk factors (lower rates of hypertension & diabetes) may work in favor. Hopefully treatment regimens will improve over time in useful ways even for resource-poor settings, but it’s hard to predict. So I completely agree with your point in the google doc that current estimates are in some sense biased toward high-capacity countries, but it’s not clear (to me) whether that would make them too high or too low overall. As you say, places with good health info are also those with high life expectancies—which means healthier but also older.
i suppose my current guess is 0.5-1% for the headline number, which is a pretty broad range but there you go. Your analysis shifted this upward a bit!
I know the Washington post opinion column isn’t the right place to post numbers, but do you have ballpark estimates for how costly (economically and/or in terms of human toll) lockdowns will be in low income and middle income countries?
I do think that some people (not saying you are one!) often underestimate the human harm of getting covid-19 in developing countries (eg, they’ll quote widely discredited numbers for IFR like .1%, which obviously is ~impossible).
So it’d be helpful to do ballpark Fermi estimates for the cost of different interventions (or not doing those interventions) vs the benefits, either for the world as a whole or a specific country in mind.
I can possibly help provide the modeling on the covid side, but I don’t have a good grasp of the “cost” side of lockdowns at the moment.
Sorry for the slow reply! I had been working on some rough estimates for total (i.e. including medium- and long-run downstream impacts) costs and benefits of e.g. lockdown vs targeted social distancing, but even in high-income countries this is hard! This paper from Layard et al (using well-being adjusted life years) is perhaps the closest I’ve seen:
http://cep.lse.ac.uk/pubs/download/occasional/op049.pdf
See also this effort for LMICs from CGD:
https://www.cgdev.org/blog/scoping-indirect-health-effects-covid-19-open-call-resources
Happy to consider collaborating on something for developing countries, if only to get a sense for which dimensions are likely to be first order and hence worthy of further study, but I’m hesitant to believe even that would be feasible with confidence. Perhaps important enough to try in any case? Also not sure I am best placed for it, as a micro economist focusing on individual behavior...
Curious to hear what you think is the best existing evidence for IFR. Indeed 0.1% seems too low overall, but for under-60s my sense was that it is probably 0.1-0.2%, as in these papers:
https://www.thelancet.com/pdfs/journals/laninf/PIIS1473-3099(20)30243-7.pdf
https://www.medrxiv.org/content/10.1101/2020.04.18.20070912v1
Of course the quality of health system resources will affect the IFR, but neither of the ones above (China, Italy) are from ideal situations either so I honestly don’t know how much worse it will be globally.
I think ~1.1% (with fairly wide uncertainty) is a fairly realistic guess for a global IFR (including all age ranges). I basically don’t buy that the balance of factors would necessarily favor poorer and younger countries over richer/healthier/older ones, though it certainly is possible.
Here’s a preliminary document listing why I believe this. Usual caveats of being a non-professional apply, and also the tone is a bit sharper than I’d use on the EA Forum (basically the intended audience was other amateur forecasters so there are certain stylistic differences, especially around caveats).
~0.1%, or even slightly lower, seems believable for <60s in some rich countries but I don’t think you want to extrapolate age-structure arguments too strongly to novel situations (in essence I think age is a biased estimator whereas something like crude death rate may not be), and if you want to look at specific countries you’d want to look at a bunch of known comorbidities*; eg per capita, Nigerians die of heart disease at ~1/10 the rate of Indians.
One thing that I didn’t mention in my document above is that even if .1%-.2% is a realistic IFR for young people in developing countries, and developing countries are skewed young, the full IFR in developing countries will likely still be much higher.
For example, Guayas province in Ecuador has had ~11,561 excess deaths from the beginning of March to mid-April (base rate is ~3000 in that time period). My understanding is that close to all of it is directly due to covid-19 (I talked to people from Ecuador and if there was mass starvations or a different epidemic that accounted for even 2x all-cause mortality I’d have heard by now). The population of Guayas is ~3 million, so this is already a lower bound of ~0.39% of the entire population(!), and I really don’t buy that anywhere near 100% of Guayas were infected as of mid-April (or more accurately late March to account for lag between infection and death).
Ecuador has a median age of 27.9, a life expectancy of 76.6, and a GDP per capita of $6400, so definitely not unusually old or unhealthy by middle-income country standards.
This is great—thanks. My belief certainly wouldn’t be that simply because of the age structure IFR is going to be lower in developing countries. I do think that will be protective, but I also think that poor health systems will (obviously) go the other way. Some risk factors (generally more stressed immune systems) may work against them; other risk factors (lower rates of hypertension & diabetes) may work in favor. Hopefully treatment regimens will improve over time in useful ways even for resource-poor settings, but it’s hard to predict. So I completely agree with your point in the google doc that current estimates are in some sense biased toward high-capacity countries, but it’s not clear (to me) whether that would make them too high or too low overall. As you say, places with good health info are also those with high life expectancies—which means healthier but also older.
i suppose my current guess is 0.5-1% for the headline number, which is a pretty broad range but there you go. Your analysis shifted this upward a bit!
so glad to see this discussion
who are you RootPi and can you reach me at ALLFED.info?