“With a Whimper: Depopulation and Longtermism” (Spears and Geruso, forthcoming), a chapter in Essays on Longtermism, strives “to bring facts from population science and population economics into dialogue with the community of longtermists who are thinking about wellbeing into the far future.”
The authors project just 30 billion future humans on the basis of worldwide declining fertility rates. Taking longtermism seriously as an ethical view, they examine some consequences: existential risk megaprojects look less affordable, technological growth looks slower, and so on.
I’m interested to hear what longtermists think about this.
The essay is well worth reading in full. From the abstract:
“To eventually achieve a flourishing far future, it is valuable that over the coming few centuries a complex global economy endures and the number of people does not become small enough to be highly vulnerable to extinction from a threat that a larger population could sustain. We review population projections and other social scientific facts that show that fertility rates that are normal in much of the world today would cause population decline that is faster and to lower levels than is commonly understood, threatening the long term future.”
I have some initial thoughts (but note that I don’t have any background in population science and I don’t work on longtermist causes):
We probably shouldn’t take 30 billion seriously as the number of future humans. I agree with Spears and Geruso that their model wouldn’t “hold until the last couple only has one child”: it’s hard to imagine eschatological Adam and Eve
But what if we did? 30 billion future humans are still more moral patients than existing humans. But there are about 30 billion existing land farm animals, and more existing aquaculture farm animals or wild animals. What this means will depend on your discount rate and interspecies moral weights
I’m more bullish than Spears and Geruso on technological developments (in e.g. in-vitro gametogenesis) mitigating effects of falling fertility rates
Even absent AGI or superintelligence, I expect artificial intelligence to take over a lot of innovation, so I think the rate of technological development could get uncoupled from population growth rate
I appreciate Spears and Geruso’s warning on public policy for population control: “Governments sometimes try to coerce people to have babies; governments sometimes try to coerce people not to have babies. It is typical, with such policies, to wreck people’s lives, wreck the economy’s human capital, and wreck society’s compact between the governed and the government”. This is important.
Summary: the number of children people are having is declining, and if we project this out for several centuries we see massive worldwide depopulation. Eventually, the population is small enough that it could get wiped out by a disaster, and most people who’ll ever live are in the past.
I find this unconvincing for several reasons:
They’re projecting a very recent phenomenon, below-replacement fertility, to last many times longer than it has so far, instead of reverting to the historical pattern.
A lot of the constraints on family size come down to cost, especially cost of housing/land, and in a depopulating world housing and land would be much cheaper.Cheap and effective birth control is very new. Historically, explicit desire for offspring didn’t matter very much: strong desire to mate was sufficient. Humanity is probably already in the process of evolving a stronger direct desire to reproduce, replacing the indirect desire. Maybe evolution is too slow (I doubt it: there’s probably a lot of variety among humans here) but I’d expect to at least see the article mention this objection?
As Michael says in his comment, technological change could easily have huge effects on the parental experience, changing his many kids people want to have.
[Disclaimer: I work at the Population Wellbeing Initiative with Mike and Dean].
As quick responses to those bullets based on my own views of this issue:
It’s true that low fertility is recent, but so is wealth and the opportunities that come with that. The main crux I’m left with is that (1) [save for Israel] there is no economically developed country that has fertility high enough to replace itself and (2) there has never been sustained depopulation via low fertility. Something is going to have to give! My money is on a non-insignificant period of depopulation.
I disagree with this point about financial costs. The cross-section and time-series evidence that richer places and times have fewer children is too hard to square with that claim, in my opinion. I’d be interested to hear if you had a further thought about why that evidence is misleading. (The opportunity costs are high, but as long as we maintain a world with good work and life opportunities for parents, I don’t see obvious reasons why this will fall).
I’d guess what they say about ‘Heritability’ on pages 14-15 might be what you’re interested in (i.e., selection pressure for sub-groups that have direct desires for children). I count myself as worried if that’s the force we’re relying on though given the universality of low fertility. (Evolution through new genetic mutations is almost certainly going to be too slow; I think Table 1 indicates births will fall to very low levels in ~300 years with European levels of fertility).
Fair enough! I don’t really know what technologies are in our future with respect to childbearing and parenting. For what it’s worth though, even if there were a (costless) artificial womb so good that it was just a button that when pressed produced a baby, most people I know who don’t have kids wouldn’t press the button. Obviously, that’s not the only tech that could change the parenting experience. But it is the most common one that’s proposed and I’m just not convinced it would move the needle much on fertility. Robo-nannies or something that made a serious dent in the time parents felt they should spend with their kids seems like a more important margin to me.
Thanks!
I wrote up a draft post, focusing on heritability; any thoughts before I publish it? I’m especially curious why you think evolution is too slow—if there’s already significant variation within humanity, and we’re in a new period than started recently with birth control + sex ed + lower taboos, it seems to me like it could be just a few generations before people with genetically higher desires for having their own children are having a large fraction of the kids?
I’d bet on that too—I just think it’s way less than 300y, and we only see modest drops before reversal.
I think my point on housing was wrong—in places with declining populations a more typical pattern is probably that less desirable areas, with the least economic opportunity, depopulating faster. So there’s still expensive housing in places where you can get good jobs, and prospective parents still face large costs if they choose to have kids.
But (a) your non-parent friends may not be the marginal parent and (b) some people would probably now press the button dozens of times.
My impression is this as well, but more as a cost issue and not a time issue. Childcare is very expensive (it’s our family’s largest expense, with three kids), and automation might be able to help with that? Not sure.
Great—glad you wrote that post up on intergenerational dynamics (and remarkably quickly!). I haven’t read through the details in a while, but I think the best paper I’ve seen trying to estimate heritability at the family-level is this one by Tom Vogl, which you might find interesting to dig into. I believe his headline finding is that in low fertility settings that this composition effect accounts for fertility rates being ~4% higher in this generation than it would otherwise be (but that’s just a refresher from my quick skim just now).
My skepticism about evolution is skepticism about the existing variance in biological preferences for children. Obviously that’s not something we can easily get at, since outcomes are the product of environment + constraints + culture + preferences, etc. But (1) this preference isn’t currently common enough to push some economically developed countries above replacement rate and (2) once social/economic conditions that generate low-fertility stabilize, this sort of mental-model would always predict increasing fertility rates (since every generations composition becomes more favorable to high-fertility). I’m not sure there’s even a single country with moderate to low fertility that’s seen an increase over the last 10-20 years, even though the demographic transition occurred in some countries a few generations ago. (And we only have a few more, ~7-10, generations worth of time until we’re at pretty low population levels).
Though I’m happy to admit that this is hard to generate convincing evidence on, so maybe in a few more generations it could start to show up in aggregate numbers. But until there’s a country or two with consistent increases in fertility, through policy or evolution or whatever, I will remain very concerned that the decline will not be self-correcting.
Don’t have much to add on the other points you made :)
Thanks! Published the post.
Much of this is just repeating things that others have said, but my initial position here is skepticism.
The model is based on a fertility trend that has arisen in a very specific cultural, economic, and technological context. I’m very skeptical that we should take it to provide any sort of reliable guide to the long term future.
It seems to me that there are plenty of ways that projecting forwards underlying trends could interrupt the fertility trend. For example, perhaps as you increase per capita wealth you get decreased child mortality, increased costs of educating children, and so on, such that having less children becomes incentivised. But if wealth continues to grow then perhaps there becomes a decoupling between economic incentives and decisions about how many children to have (because marginal wealth becomes less important so costs in terms of marginal wealth matter much less in terms of their impact on utility).
Low fertility itself seems likely to lead to cultural changes. I feel pretty sceptical of the idea that we end up in a world with a radically shrinking population where we can carry forward the trends that are familiar from a world with a growing population.
AI could easily change the connection between population growth and innovation, in a way that means progress could continue absent population growth (and this progress will plausibly itself gives the tools necessary to resolve issues of population if we become worried).
AI might itself count as population in whatever sense matters.
Fertility technologies might change how easy it is to have children and might lead to a decoupling between parental choice and societal birthrate (for example, you could imagine a world of artificial wombs where the government is responsible for creating the next generation, and where children are co-raised by society; clearly there might be issues with such a world, but the fertility rate itself is not the issue).
I believe that evolutionary pressures tend to push genes responsible for fertility to evolve to fixity. The genes now responsible for fertility are increasingly those related to wanting children. We should expect these genes to evolve to fixity.
One might retreat to saying we should have a small credence in the relevant models but claiming this suffices to justify action. I’m skeptical that we should even have a sufficiently high (small) credence for this argument to go through. Projecting this population trend forwards 300 years through the radical change we should expect over that time seems to me not very informative.
I recognise that the people working on this are better informed than me on this topic, and that seems like a relevant consideration. But I worry this is… kinda EA nerdbait. Clever big picture thinking, backed by quantitative models, revealing a hidden catastrophe that others have not foreseen sufficiently clearly. I’m not saying such things never get at the truth, but I do think it’s reasonable to approach them with an initial attitude of skepticism, even in the face of the existence of enthusiastic proponents.
Many or most longtermists take artificial conscious minds seriously and as morally important, and there could be many orders of magnitude more of those in the future.
Also, even just with biological humans, we could dramatically lower the costs of having and raising children by making large parts of the process artificial, as you point out (artificial wombs, AI parents or no parents at all and more communal raising of children), allowing far more humans to exist.