The techno-pessimists are not our enemies. It’s wrong to think of techno-optimists and techno-pessimists as two opposing teams, each arguing for its own point of view; in fact, we’re on the same side. Both want faster technological progress, we are merely debating how hard the challenge will be. In fact, it’s not even clear that we want substantively different policies.
I had thought that the ancient life expectancy you describe would apply to men, but that women's life expectancy was much lower due to death from childbirth. One of the modern miracles is the drop in infant/child mortality, but also in reducing the death from childbirth for women. This is an area where the US lags behind and contributes to our lower life expectancy relative to other developed countries.
Generally agree, not deserving of pessimism, but it is difficult to sort out tech impacts here.
Cleaner power and cars will reduce local pollution, providing bankshot health wins. I'm definitely bullish on medical therapies too.
The problem is that behavioral factors can swamp everything else going on. Reductions in smoking, as you mention, contributed a lot to the decline. (Not only in cancer and heart attack rates, but even in things like household fires: https://www.nber.org/papers/w16625 )
In the other direction, that dip a year or so ago was mostly from spiking suicide rates in the mountain west, which doesn't seem to really be a tech story (maybe the lack of mental health tech?) Also, cheap synthetic opiods, which could be called tech working against us.
I'm still holding out for a moonshot on malaria or cancer to improve global health.
Along the way I'm expecting non-health tech to contribute incremental improvements too.
But I still worry the overall trendline will be dominated by behavioral choices, which still need clever new policy solutions in areas where there aren't easy answers.
As an about to be elderly person (68 isn't old, I tell you), quite a while ago I realized that "life expectancy" itself wasn't the be all/end all of statistics. What I want to know is life expectancy at age {whatever I am at the time}. But I think that LE at {some appropriate age} is way more informative than LE at birth. (Maybe even 2 or 3 numbers: LE65 and LE80 would tell you how retirees overall do and how the healthy elderly do.) The US has the worst infant mortality rate in the developed world by a lot, so there's still a lot of low-hanging fruit in the LE at birth number.
FWIW, I'm a TO in the long run, TP in the short run. It's hard to replace an established technology. DRAM and Li batteries keep getting better and cheaper and keep getting harder to replace. Also, we tend to hyperventilate over things that don't have any there there: the current AI boom is a perfect example. (Long story short: AI has been ignoring the basics and depending on statistical and other dumb calculations: see "Rebooting AI" for the painful details.)
In the long run, there are quite a few people in AI who get it that current programs can't do simple everyday reasoning and that there's interesting work to be done there. And the long, painful grind of figuring out the incredible complexities of biological systems has proved it's worth just last year with mRNA vaccines.
As for infant mortality, we could probably take it down from 5.8 to 3.4 or so. That would be great, but the gains in headline life expectancy would be small.
Ah, you're probably right there. Japan is at 1.8 deaths per 1000 births, and that's "only" three times better than the US, so while it's a horrific distance, we're not going to get close without major social changes, and an extra 2 or 4 per thousand people living a full life won't change numbers much. Which brings up another point here (differences between demographic groups): there was a big deal about "deaths of despair", but not only was that a tiny effect, it might even have been a statistical error (as discussed on A. Gelman's blog). In particular, the demographic subgroup of concern has life expectancies far far greater than those of minorities and the poor in the US. And that difference also applies to infant mortality as well.
Yeah, this is like the math thing where if you walk 1 mph faster when you're not moving at all, you decrease your travel time infinitely, but if you're already walking at 5 mph then walking 1 mph faster only decreases your travel time by a small amount...
If you are a company making a battery purchase, one important question is the lifetime of the proposed battery. If it's a new technology, you just don't know. Meanwhile, people are putting a lot of effort into twiddling existing technologies.
Oops, I'm being unclear: by "replace" I meant "replace with a new technology". My point is that if Xfrob Technologies has a new battery technology (and there are a bunch in the works), it's hard to beat Li, despite Li's problems. If UserProducts, Inc. is putting batteries in a product that needs to last 7 years (hooray for consumer protection laws!), you know what 7-year old Li batteries are like, and can plan for it. If the new technology doesn't have 7 years of actual experience, you don't know what will happen and probably tell Xfrob to take a walk. Li (last I checked) wasn't great for car batteries, but you (you, not me; I don't own a car) don't want to buy an electric car that needs new batteries every 3 years. Accelerated lifetime testing can help, but it's an estimate, not experience.
> The US has the worst infant mortality rate in the developed world by a lot
It really has two infant mortality rates. 4 point something for most of the population, right in line with Canada/UK/Australia/Taiwan. Then 10.8 for African Americans, who we are absolutely failing.
It's bimodal, both a bit better and so much worse than it looks.
Thanks for the correct numbers. I was aware of the bimodal bit (see my note below), but "four point something" is over twice Japan's rate. so we could do better even there. (Hmm. There are only 10 countries at 3.0 and lower, so maybe I'm being too optimistic.)
Sorry, didn't mean to nitpick, totally agree with you that much more progress can be made. I just footstomp this to encourage folks to think hard about identifying inequities in care and coming up with targeted services for specific groups, rather than just relying on broad programs hoping that will be enough. I think Emily Oster has written about this as involving maternal information gaps a lot. If we're choosing between a building a new hospital ward with the latest equipment, or an ad campaign summarizing what we know about things like SIDS and FASD and folate, maybe the ads would be more really effective and much lower cost.
Up to this point our Big Research (tm) hasn't treated aging itself and senescence as an actual disease, but rather a natural process we don't need to be spending $$ on. If that were to change these graphs may become significantly different.
My understanding is that the drop in US life expectancy is largely related to increases in poverty since trickle-down Reaganomics took hold in the '80's, obesity-related diseases such as diabetes and heart disease which are closely correlated with poverty (and poor federal nutrition guidelines related to fat and sugar), and deaths of despair such as overdoses and alcoholism, which are driven by the decline in manufacturing due to free trade agreements.
I.e. *none* of the drop in US life expectancy is even vaguely related to technology.
"We’ve picked the low-hanging fruit of science" Weird argument! The whole techno-optimist argument is that we now have the tools(such as the Cryo-EM "resolution revolution", Crispr, progress in AI etc. etc.) to tackle the difficult stuff.
It's true in a limited sense. Most of the experiments that people could do with things gathered from their garage and kitchen have been done. Special equipment is needed to make new discoveries, and it tends to be expensive. This is why we are hearing some questions about how best to fund science and how to mix daring, long-odds projects with those that seem likely to succeed.
One minor quibble: where it states that female life expectancy accelerated between 1930-1950 because of improvements to maternal mortality. The data is for 45 year olds. I don’t think there was substantial enough maternity (aka childbirth) in this cohort to make much difference. One could posit some correlation with the Great Depression though why work would make LE go up...
Interesting and valid comments about life expectancy. However, I do not think that life expectancy is a valid proxy to measure life expectancy with. Not at all.
I'm not sure I buy the technology/lifestyle factors distinction. If, for example, vaping is a much safer way to ingest nicotine, then people might smoke less due to being able to substitute with vapes. Would that be technology or lifestyle? Similarly low-fat yoghurt or sweetened soft drinks may also be things that blur this line.
I suspect there is a complicated dynamic that plays out when it comes to life expectancy. In a welfare state, life expectancy improvements will put an undue burden on the exchequer over time. That burden should ideally be transferred to a growing working-age population through a progressive, but moderate tax system. In the absence of that, people will be forced to under consume, in anticipation of an inadequate welfare system. This is a negative incentive for technological progress. The alternative is that their productivity gains need to be overtaxed, which is also a negative incentive, or welfare states borrow heavily, which transfers the burden elsewhere in the world and delays the problem but doesn't really fix it. Ultimately, if life expectancy improvements flatten out, without an increase in the working-age population, technological progress will be stymied (ceteris paribus).
I had thought that the ancient life expectancy you describe would apply to men, but that women's life expectancy was much lower due to death from childbirth. One of the modern miracles is the drop in infant/child mortality, but also in reducing the death from childbirth for women. This is an area where the US lags behind and contributes to our lower life expectancy relative to other developed countries.
Generally agree, not deserving of pessimism, but it is difficult to sort out tech impacts here.
Cleaner power and cars will reduce local pollution, providing bankshot health wins. I'm definitely bullish on medical therapies too.
The problem is that behavioral factors can swamp everything else going on. Reductions in smoking, as you mention, contributed a lot to the decline. (Not only in cancer and heart attack rates, but even in things like household fires: https://www.nber.org/papers/w16625 )
In the other direction, that dip a year or so ago was mostly from spiking suicide rates in the mountain west, which doesn't seem to really be a tech story (maybe the lack of mental health tech?) Also, cheap synthetic opiods, which could be called tech working against us.
I'm still holding out for a moonshot on malaria or cancer to improve global health.
Along the way I'm expecting non-health tech to contribute incremental improvements too.
But I still worry the overall trendline will be dominated by behavioral choices, which still need clever new policy solutions in areas where there aren't easy answers.
Yeah.
Things I'm curious about on the medical/behavioral front are VR, and also controlled use of psychedelics as mental health treatment. 🤔
As an about to be elderly person (68 isn't old, I tell you), quite a while ago I realized that "life expectancy" itself wasn't the be all/end all of statistics. What I want to know is life expectancy at age {whatever I am at the time}. But I think that LE at {some appropriate age} is way more informative than LE at birth. (Maybe even 2 or 3 numbers: LE65 and LE80 would tell you how retirees overall do and how the healthy elderly do.) The US has the worst infant mortality rate in the developed world by a lot, so there's still a lot of low-hanging fruit in the LE at birth number.
FWIW, I'm a TO in the long run, TP in the short run. It's hard to replace an established technology. DRAM and Li batteries keep getting better and cheaper and keep getting harder to replace. Also, we tend to hyperventilate over things that don't have any there there: the current AI boom is a perfect example. (Long story short: AI has been ignoring the basics and depending on statistical and other dumb calculations: see "Rebooting AI" for the painful details.)
In the long run, there are quite a few people in AI who get it that current programs can't do simple everyday reasoning and that there's interesting work to be done there. And the long, painful grind of figuring out the incredible complexities of biological systems has proved it's worth just last year with mRNA vaccines.
DJL in Tokyo.
As for infant mortality, we could probably take it down from 5.8 to 3.4 or so. That would be great, but the gains in headline life expectancy would be small.
Ah, you're probably right there. Japan is at 1.8 deaths per 1000 births, and that's "only" three times better than the US, so while it's a horrific distance, we're not going to get close without major social changes, and an extra 2 or 4 per thousand people living a full life won't change numbers much. Which brings up another point here (differences between demographic groups): there was a big deal about "deaths of despair", but not only was that a tiny effect, it might even have been a statistical error (as discussed on A. Gelman's blog). In particular, the demographic subgroup of concern has life expectancies far far greater than those of minorities and the poor in the US. And that difference also applies to infant mortality as well.
Yeah, this is like the math thing where if you walk 1 mph faster when you're not moving at all, you decrease your travel time infinitely, but if you're already walking at 5 mph then walking 1 mph faster only decreases your travel time by a small amount...
Why do you think Li batteries keep getting harder to replace?
If you are a company making a battery purchase, one important question is the lifetime of the proposed battery. If it's a new technology, you just don't know. Meanwhile, people are putting a lot of effort into twiddling existing technologies.
But we see Li-ion battery demand climbing steadily, don't we?
Oops, I'm being unclear: by "replace" I meant "replace with a new technology". My point is that if Xfrob Technologies has a new battery technology (and there are a bunch in the works), it's hard to beat Li, despite Li's problems. If UserProducts, Inc. is putting batteries in a product that needs to last 7 years (hooray for consumer protection laws!), you know what 7-year old Li batteries are like, and can plan for it. If the new technology doesn't have 7 years of actual experience, you don't know what will happen and probably tell Xfrob to take a walk. Li (last I checked) wasn't great for car batteries, but you (you, not me; I don't own a car) don't want to buy an electric car that needs new batteries every 3 years. Accelerated lifetime testing can help, but it's an estimate, not experience.
Gotcha.
Great, informative stuff. Thank you for your posts. 👍
Guy in Touhoku
> The US has the worst infant mortality rate in the developed world by a lot
It really has two infant mortality rates. 4 point something for most of the population, right in line with Canada/UK/Australia/Taiwan. Then 10.8 for African Americans, who we are absolutely failing.
It's bimodal, both a bit better and so much worse than it looks.
Thanks for the correct numbers. I was aware of the bimodal bit (see my note below), but "four point something" is over twice Japan's rate. so we could do better even there. (Hmm. There are only 10 countries at 3.0 and lower, so maybe I'm being too optimistic.)
Sorry, didn't mean to nitpick, totally agree with you that much more progress can be made. I just footstomp this to encourage folks to think hard about identifying inequities in care and coming up with targeted services for specific groups, rather than just relying on broad programs hoping that will be enough. I think Emily Oster has written about this as involving maternal information gaps a lot. If we're choosing between a building a new hospital ward with the latest equipment, or an ad campaign summarizing what we know about things like SIDS and FASD and folate, maybe the ads would be more really effective and much lower cost.
Up to this point our Big Research (tm) hasn't treated aging itself and senescence as an actual disease, but rather a natural process we don't need to be spending $$ on. If that were to change these graphs may become significantly different.
My understanding is that the drop in US life expectancy is largely related to increases in poverty since trickle-down Reaganomics took hold in the '80's, obesity-related diseases such as diabetes and heart disease which are closely correlated with poverty (and poor federal nutrition guidelines related to fat and sugar), and deaths of despair such as overdoses and alcoholism, which are driven by the decline in manufacturing due to free trade agreements.
I.e. *none* of the drop in US life expectancy is even vaguely related to technology.
"We’ve picked the low-hanging fruit of science" Weird argument! The whole techno-optimist argument is that we now have the tools(such as the Cryo-EM "resolution revolution", Crispr, progress in AI etc. etc.) to tackle the difficult stuff.
Also: The recent(accelerated) decline of cancer deaths was partly attributed to rise of immunotherapy. https://www.healthline.com/health-news/cancer-deaths-are-down-30-percent Seems like a better measurement of progress in biotech to me.
It's true in a limited sense. Most of the experiments that people could do with things gathered from their garage and kitchen have been done. Special equipment is needed to make new discoveries, and it tends to be expensive. This is why we are hearing some questions about how best to fund science and how to mix daring, long-odds projects with those that seem likely to succeed.
Nice post.
One minor quibble: where it states that female life expectancy accelerated between 1930-1950 because of improvements to maternal mortality. The data is for 45 year olds. I don’t think there was substantial enough maternity (aka childbirth) in this cohort to make much difference. One could posit some correlation with the Great Depression though why work would make LE go up...
Interesting and valid comments about life expectancy. However, I do not think that life expectancy is a valid proxy to measure life expectancy with. Not at all.
On to part two, thanks, this is good.
I'm not sure I buy the technology/lifestyle factors distinction. If, for example, vaping is a much safer way to ingest nicotine, then people might smoke less due to being able to substitute with vapes. Would that be technology or lifestyle? Similarly low-fat yoghurt or sweetened soft drinks may also be things that blur this line.
I suspect there is a complicated dynamic that plays out when it comes to life expectancy. In a welfare state, life expectancy improvements will put an undue burden on the exchequer over time. That burden should ideally be transferred to a growing working-age population through a progressive, but moderate tax system. In the absence of that, people will be forced to under consume, in anticipation of an inadequate welfare system. This is a negative incentive for technological progress. The alternative is that their productivity gains need to be overtaxed, which is also a negative incentive, or welfare states borrow heavily, which transfers the burden elsewhere in the world and delays the problem but doesn't really fix it. Ultimately, if life expectancy improvements flatten out, without an increase in the working-age population, technological progress will be stymied (ceteris paribus).
Forget technology to increase lifespan: I'm just going to eat more 寿司. 🍣🍙👍