It makes sense to me that the DALYs don't include the economic side of the wage loss, but given how well-evidenced the wage loss is (and how strongly that implies cognitive or energy loss), it's surprising that there's no accounting for the cognitive effects.
I agree that DALYs don’t (and shouldn’t) capture everything, and that well-calculated DALY burden should provide a lower bound rather than the best guess estimate - things like income are obviously quite important, but not captured in DALYs.
GBD distinguishes causes from risk factors to prevent double counting. To avoid this double counting issue, a constraint was introduced where cause-specific estimates had to sum to real population totals.
Correct (and probably would have been worth mentioning in the post - thanks for pointing it out!). Causes are indeed designed to avoid double-counting. Risk factors don’t sum / are designed to double-count. For example, child growth failure is 79.4 million DALYs, stunting is 33 M, wasting is 39.2 M, childhood underweight is 52.2 M - they sum to well over the total of 79.4.
The argument I’m making (perhaps more clearly in the spreadsheet than the post) is that stunting is a risk factor for many causes it’s not currently being tallied as a risk factor for (e.g. cognitive effects). A reasonable corollary from this could be that GBD as a whole is under-counting cognitive effects, but it could also be the case that they are being counted correctly, but not attributed to the risk factor of stunting - even though they should be since risk factors, unlike causes, are overlapping.
This matters because the cognitive and physiological harms captured in those two disability weights are the same harms that already accrue to a stunted child's own anemia and cognitive-impairment burden elsewhere in GBD's accounting. You're supposed to be able to sum DALYs averted. But if you do it your way, a CEA can't count both DALYs averted from reduced stunting and separately DALYs averted from reduced anemia or cognitive impairment.
Many stunted people later in life are not anemic, my argument is not that we should attribute the DALYs from anemia to stunting, but rather that stunting probably acts as a risk factor for health-state burdens fairly similar to those caused by anemia (well, that’s my upper bound - I think probably more like half that burden). If a stunted person continues to not receive sufficient nutrition and is anemic later in life, I think the anemia is being counted correctly but contend there is additional burden not counted.
The wage effects of stunting are well documented, which to me is pretty strong evidence for cognitive or energy-level health burden. However, as I understand it, that's currently not counted in GBD. Browsing Vizhub (https://vizhub.healthdata.org/gbd-compare/), the only intellectual disability listed is environmental/occupational and labeled idiopathic, meaning the cognitive effects of malnutrition are not modeled.
So my claim is not that stunting should take a larger chunk of the DALYs in GBD from other risk factors, but rather that the overall number of DALYs listed in GBD is too low due to failure to price in these long-term effects.
I recently wrote a biofortification related post, and realized it was probably worth looking through the forum to see what else the community has done in that space.
To my surprise, this was the only other hit - "running an information campaign on biofortification" as a promising idea.
Did you ever run that campaign, or is there any writeup of why you thought it was promising, results, etc? I'm realizing this may be a much more neglected area than I had thought. I think most EAs are familiar with at least golden rice, thus the surprise that nobody seems to have written about it (or at least not using the widely-used technical term).
AI companies are not transparent enough about the environmental cost of generating images, so it is possible that I should stop using these tools.
I wouldn't worry much about this, those costs on a per-user basis are incredibly low, almost certainly much lower than other things you won't worry about. Andy Masley is a great source on the topic: https://blog.andymasley.com/p/a-cheat-sheet-for-conversations-about
That said, many (most?) parts of society have a strong anti-AI preference, and strategically it doesn't make sense to anger your audience.
I think it is worth separating "don't do this because it's morally wrong" from "don't do this because it's not effective for winning hearts and minds".
Interesting post, certainly an interesting comparison and an existence proof that technology that's somewhat difficult to create but trivial to distribute and reproduce can be regulated to oblivion in large sections of the world for decades.
A point conspicuous by its absence: the overregulation of GM crops was (and remains) a mistake, or at the least is nearly universally agreed to be so by the people with the most technical knowledge (e.g. people with PhDs in plant biology).
I understand whether it was wise to grossly curb deployment of GM crops was not the point of the post, merely whether it was politically feasible starting with a relatively small contingent of protesters. I'm still miffed that the anti-science and overall almost certainly negative-EV-ness of the GM overregulation wasn't mentioned, especially given quotes like "This coincided with key 'trigger events' like Mad Cow Disease and the arrival of GM crops in March 1996" which would suggest a causal connection between the two.
Glad you enjoyed!
It makes sense to me that the DALYs don't include the economic side of the wage loss, but given how well-evidenced the wage loss is (and how strongly that implies cognitive or energy loss), it's surprising that there's no accounting for the cognitive effects.
Thanks for the detailed comment Rey!
I agree that DALYs don’t (and shouldn’t) capture everything, and that well-calculated DALY burden should provide a lower bound rather than the best guess estimate - things like income are obviously quite important, but not captured in DALYs.
Correct (and probably would have been worth mentioning in the post - thanks for pointing it out!). Causes are indeed designed to avoid double-counting. Risk factors don’t sum / are designed to double-count. For example, child growth failure is 79.4 million DALYs, stunting is 33 M, wasting is 39.2 M, childhood underweight is 52.2 M - they sum to well over the total of 79.4.
The argument I’m making (perhaps more clearly in the spreadsheet than the post) is that stunting is a risk factor for many causes it’s not currently being tallied as a risk factor for (e.g. cognitive effects). A reasonable corollary from this could be that GBD as a whole is under-counting cognitive effects, but it could also be the case that they are being counted correctly, but not attributed to the risk factor of stunting - even though they should be since risk factors, unlike causes, are overlapping.
Many stunted people later in life are not anemic, my argument is not that we should attribute the DALYs from anemia to stunting, but rather that stunting probably acts as a risk factor for health-state burdens fairly similar to those caused by anemia (well, that’s my upper bound - I think probably more like half that burden). If a stunted person continues to not receive sufficient nutrition and is anemic later in life, I think the anemia is being counted correctly but contend there is additional burden not counted.
The wage effects of stunting are well documented, which to me is pretty strong evidence for cognitive or energy-level health burden. However, as I understand it, that's currently not counted in GBD. Browsing Vizhub (https://vizhub.healthdata.org/gbd-compare/), the only intellectual disability listed is environmental/occupational and labeled idiopathic, meaning the cognitive effects of malnutrition are not modeled.
So my claim is not that stunting should take a larger chunk of the DALYs in GBD from other risk factors, but rather that the overall number of DALYs listed in GBD is too low due to failure to price in these long-term effects.
I recently wrote a biofortification related post, and realized it was probably worth looking through the forum to see what else the community has done in that space.
To my surprise, this was the only other hit - "running an information campaign on biofortification" as a promising idea.
Did you ever run that campaign, or is there any writeup of why you thought it was promising, results, etc? I'm realizing this may be a much more neglected area than I had thought. I think most EAs are familiar with at least golden rice, thus the surprise that nobody seems to have written about it (or at least not using the widely-used technical term).
I wouldn't worry much about this, those costs on a per-user basis are incredibly low, almost certainly much lower than other things you won't worry about. Andy Masley is a great source on the topic: https://blog.andymasley.com/p/a-cheat-sheet-for-conversations-about
That said, many (most?) parts of society have a strong anti-AI preference, and strategically it doesn't make sense to anger your audience.
I think it is worth separating "don't do this because it's morally wrong" from "don't do this because it's not effective for winning hearts and minds".
Interesting post, certainly an interesting comparison and an existence proof that technology that's somewhat difficult to create but trivial to distribute and reproduce can be regulated to oblivion in large sections of the world for decades.
A point conspicuous by its absence: the overregulation of GM crops was (and remains) a mistake, or at the least is nearly universally agreed to be so by the people with the most technical knowledge (e.g. people with PhDs in plant biology).
I understand whether it was wise to grossly curb deployment of GM crops was not the point of the post, merely whether it was politically feasible starting with a relatively small contingent of protesters. I'm still miffed that the anti-science and overall almost certainly negative-EV-ness of the GM overregulation wasn't mentioned, especially given quotes like "This coincided with key 'trigger events' like Mad Cow Disease and the arrival of GM crops in March 1996" which would suggest a causal connection between the two.