I have some sad news to report: My podcast, Econ 102, is going on indefinite hiatus. Erik Torenberg, my excellent co-host, is very busy at his new job at a16z, and we weren’t able to keep up a regular schedule of podcasting. I apologize to all the fans of Econ 102!
Time permitting, Erik and I may return with more content later. And in the meantime, I am actively thinking about some other ways to deliver you voice and video content, so stay tuned.
In the meantime, here’s this week’s roundup of interesting econ-related stuff!
1. Another “L” for basic income
I had high hopes for the idea that just giving people cash would fix a lot of society’s problems. I still think a system of unconditional cash benefits would be simpler, fairer, and easier to navigate than many of our current welfare programs, and I still think it’s worth giving poor people money in order to make them less poor. But over the past few years, a bunch of new evidence has shown that the costs of cash giveaways are higher (in terms of incentivizing people to stop working), and the social benefits are much narrower, than boosters like myself had believed. Kelsey Piper had a great writeup of this disappointing evidence a few months ago, and I wrote up some thoughts in one of my earlier roundups.
Now we have another piece of evidence showing that cash benefits solve fewer problems than we’d like it to solve. Aaltonen, Kaila, and Nix (2025) study a recent basic income experiment in Finland. In 2017, 2000 unemployed Finnish people were randomly selected to get a change in their welfare benefits. Instead of Finland’s usual unemployment benefits, the 2000 lucky people got 2 years of cash — about $658 per month at today’s exchange rates. This allowed them to A) keep receiving cash even if they started working, and B) avoid the normal job-search requirement that comes along with unemployment benefits.
The authors find that the basic income was effective in terms of raising people’s incomes. This isn’t surprising — if you don’t stop giving unemployed people cash when they get a job, they’re going to get more money overall.
Unfortunately, though, there wasn’t much impact on labor income relative to the control group. We’d hope that if we gave people money unconditionally instead of yanking it away the moment they got a job, it would incentivize them to go find work. But in this experiment, people who no longer faced that benefit cutoff were no more likely to go make money in the market:
That’s disappointing. But what’s also disappointing is the main result of the paper: People who received the basic income were no less likely to commit crimes. Aaltonen et al. write:
[W]e estimate the impact of being randomized to receive a basic income on crime perpetration. We find no effect on whether treated individuals perpetrate crimes. In the two years following the start of the experiment, individuals in the treatment group were statistically insignificantly 0.5 percentage points more likely to be suspected of a crime…representing a 2 percent increase relative to the control group mean of 20 percent…
We also find that the introduction of basic income had a statistically insignificant impact on the probability of being charged [with serious crimes] in district court…Over the two-year follow-up period, the point estimate suggests that the basic income experiment increased the probability of being charged by an insignificant 0.2 percentage points…corresponding to a 4 percent increase relative to the control group…
The results [also] show no evidence that treated individuals are more likely to engage in disorderly conduct, suggesting that the basic income intervention also does not increase lower-level criminal activity…
[W]e [also] find no evidence that introducing a basic income altered victimization risk…We also find no effects when we look separately by crime type.
In other words, giving people basic income instead of traditional welfare doesn’t seem to make them less criminal, and doesn’t seem to make them safer from crime. (If anything, it made people slightly more criminal, though the result wasn’t statistically significant, so “not noticeably different from zero” is the safest takeaway here.)
In other words, basic income has taken yet another “L” here. We’d like to tell ourselves that poverty is the root of crime, but in the short term, that’s not the case — giving people more money doesn’t make them less criminal, at least in Finland. The root causes of crime are either longer-term economic factors, or deeper sociological factors.
Cash benefits still give out cash, which makes people less poor. But they don’t have a lot of the side benefits many of us had hoped for. A lot of what happens in society can’t easily be reduced to how much money people make.
2. The Mississippi Miracle still doesn’t look like a myth
A couple of years ago, I wrote about Mississippi apparently achieving a breakthrough in teaching kids to read. From what I could tell, the improvement in fourth-grade reading scores was attributable to two things: improved teaching methods for reading, and a policy of holding kids back and making them repeat grades if they couldn’t read. I wrote:
I think there are basically two lessons we can learn here. First, older education techniques are sometimes better than newer ones, and deserve to be brought back (here is another example). Second, the progressive idea of giving struggling students more resources and the conservative idea of holding students back until they’re proficient at testable skills end up working very well together.
A few days ago, Andrew Gelman — generally one of my favorite debunkers of bad statistics — wrote a blog post about this so-called Mississippi Miracle, in which he chided me for not thinking about the selection effect of holding the worst-performing kids back:
[I]f you don’t let them pass to a higher grade, you’re gonna see higher average scores among the students who do take the test. This is something that an economics journalist should realize!
Well, yeah. I admit that I didn’t realize that holding some kids back could introduce a selection effect! I simply assumed that those held-back kids would take the test eventually, when they reached fourth grade, since A) every fourth-grader has to take the NAEP, and B) America doesn’t allow kids to drop out of school before high school.
So yes, I just assumed that the held-back kids were still in the sample of test-takers, and that holding more kids back wouldn’t distort the test results in Mississippi.
In fact, I still think I made the right assumption. Where does Gelman get the idea that some of the kids are being dropped from the sample? He gets it from a recent article by Wainer, Grabovsky, and Robinson that makes the same claim:
But it was the second component of the Mississippi Miracle, a new retention policy…that is likely to be the key to their success. Third-graders who fail to meet reading standards are forced to repeat the third grade. Prior to 2013, a higher percentage of third-graders moved on to the fourth grade and took the NAEP fourth-grade reading test. After 2013, only those students who did well enough in reading moved on to the fourth grade and took the test…It is a fact of arithmetic that the mean score of any data set always increases if you delete some of the lowest scores.
But why do Wainer et al. think the students who are held back end up being dropped from the sample of test-takers, instead of merely delayed to another year’s sample? It’s not clear. Kelsey Piper thinks Wainer et al. just haven’t thought very carefully about how education works in America:
[A] major plank of Mississippi’s reading reforms is a test of basic reading fluency administered at the end of third grade. Kids who don’t pass it are held back a year. For at least five years, people have argued that maybe this, rather than any actual teaching skill present in Mississippi, is what’s driving the state’s improvements.
A provocative new article from Howard Wainer, Irina Grabovsky, and Daniel H. Robinson in Significance argued that, in fact, nearly all of Mississippi’s results are driven by the third-grade retention policy, not by the phonics instruction, curriculum changes, or the teacher training that accompanied them. It has gone viral, with lots of glee in certain quarters, where it was sometimes taken as proof that there’s nothing other states need to learn from Mississippi after all…This is an important debate, but I’ve been dismayed to see their article treated as a significant contribution to it…
Strangely, [Wainer et al.] treated holding back 5% of students as identical to truncating the lowest 5% of test scores. But those are two different things, which makes their conclusion, that truncating the scores would be sufficient to explain Mississippi’s gains and therefore that other reforms played no role,invalid…
A student that repeats the third grade does not conveniently vanish off the face of the earth. They just … take third grade again, and then they move on to fourth grade. The state still tests them; they just do so a year later…If a student is held back a year, they still take the test again when they do reach fourth grade…Under Mississippi’s retention law, a student can usually only be retained for a maximum of two years.
They will still go on to fourth grade and therefore still take the test. This makes Wainer et al.’s entire analysis of how the mean shifts in a truncated dataset void.
What’s more, Piper digs into the Mississippi test score data — something that Gelman admits he didn’t do — and finds some very clear signs that the test score gains aren’t from holding kids back:
If all of Mississippi’s gains were from the bottom 10% of students being held back and then getting to take the test a year later…We would not expect the strongest students to be affected at all — none of them are held back, and almost no child is going to go from bottom 10th to top 10th in a single year…The NAEP publishes test scores broken down by decile. And what we see is that Mississippi has seen gains in every decile. [emphasis mine]
Now of course, switching to a policy of holding more kids back could temporarily juice scores for a couple of years, while the first cohort of weaker kids was being held back. But Piper notes that a lot of Mississippi’s test score gains came from well before 2015, when the policy switched. And she also notes that the gains haven’t been reversed since 2017, meaning that any illusory gains from that first held-back cohort are out of the sample by now.
Anyway, Andrew Gelman did apparently realize that the kids who are held back aren’t just dropped from the sample. At the end of his post, he writes:
I still wonder what happens with those kids who are held back and are then tested a year later. I guess they improve on average a lot on their own, no matter what is done, during that year.
But the kids who are held back are not “on their own”. They are still in school. And to claim that they improve “no matter what is done” doesn’t make sense, since all of them are still in school. So I’m not sure what Gelman is talking about here.
I’m not sure why Gelman, who is normally among the most perceptive thinkers when it comes to data issues like this, simply takes the conclusions of Wainer et al. at face value instead of interrogating their extremely questionable assumptions.1 Being a researcher instead of merely a journalist, he should dig into the data himself, and investigate whether there’s actually any reason to believe that the Mississippi Miracle is fake.
3. Someone out there is boosting Nick Fuentes
In the aftermath of the death of right-wing activist Charlie Kirk, there has been a lot of consternation that his shoes would be filled by Nick Fuentes. Fuentes is the leader of a movement called the Groypers that emphasizes antisemitism, ethno-nationalism, and various other race-war stuff. Richard Hanania has done a great job tracking how Groyperism has slowly taken over among the Republican staffer class, and among young Republican circles more generally. Fuentes is at the head of that movement, with his podcast briefly hitting the #1 spot on Spotify before being taken down.
A new study by the Network Contagion Research Institute has shown that Fuentes is being boosted by some shadowy, nefarious forces. Colin Wright, himself a man of the political right, has a good post explaining the findings:
According to NCRI, Fuentes’s apparent rise was driven by coordinated manipulation of online platforms, artificial engagement meant to boost his posts, and an information ecosystem in which major media outlets can be misled into thinking a fringe figure is suddenly influential…
The report’s most shocking finding is just how wildly Fuentes’s engagement numbers differ from those of other political influencers. NCRI compared the first 30 minutes of engagement on 20 of his recent posts with those from four major online figures—Elon Musk, Hasan Piker, Steven “Destiny” Bonnell, and Ian Carroll. Incredibly, Fuentes outperformed all of them in early retweets, including Musk, whose follower count is over 200 times higher…
None of this makes sense if the engagement is organic. According to NCRI’s report, this is explained by the fact that 61 percent of Fuentes’s early retweets come from accounts that repeatedly retweeted several of his posts within the same 30-minute window. This is not what you’d expect if these were random users scrolling their feeds. Rather, these accounts appear to be waiting for Fuentes to post so they could amplify his content almost instantly…
When NCRI dug into who these accounts actually were, 92 percent were completely anonymous. No real names, no real photos, no location, no identifiable personal information…When they examined Fuentes’s most viral posts—three from before the assassination of Charlie Kirk and three after—[NCRI] found that nearly half of all retweets came from foreign accounts, heavily concentrated in India, Pakistan, Nigeria, Malaysia, and Indonesia.
It’s not clear who is boosting Nick Fuentes. Pakistan, Nigeria, and so on are simply places where you can hire a bunch of people (and bots) to do artificial engagement-boosting. It could be any one of America’s foreign adversaries (all of whom Fuentes praises on his show), or it could be rightist rich people in America trying to boost their extremist ideology.
Either way, the proper takeaway here is NOT “Hahaha Fuentes’ popularity is fake, we don’t have to worry about Groypers taking over.” Fuentes’ popularity started out fake, but it didn’t entirely stay there — he got an interview with Tucker and has been covered by all the major media outlets, and his podcast soared in popularity (even if some of those downloads might be from bots). His ideas have probably been able to influence a significant number of young conservatives.
The proper lesson here is that new media technology — especially X and other viral media platforms — has tipped the scales of political discourse toward the worst extremists in our society. Unless and until we develop new institutions capable of gatekeeping out people like Fuentes, shadowy actors will use bots and paid engagement farming to boost them far out of proportion to whatever natural appeal their ideas would have had — thus destabilizing our democratic society.
4. Good news about macrosociology
We hear a lot about worsening social trends in America — higher murder rates, more suicide, more drug overdoses, more traffic deaths, and so on. But it’s starting to look like a lot of that deterioration was just a temporary bump from the pandemic and/or the nationwide unrest in 2020. Jeremy Horpedahl recently posted a great chart showing that most of these trends have been reversing since 2022:
If these trends were due to specific policies — gun control, health care policy, drug enforcement, road safety policy — we wouldn’t expect them to line up so perfectly, and we probably would expect them to line up with changes in presidential administration or Congressional control. But they do line up, and the timing doesn’t seem to fit any political changes.
In other words, it seems like it’s possible for societies to just “break” and “fix themselves” in ways we don’t entirely understand. In other words, there is some sort of sociological macro-cycle out there.
Macrosociology?
5. Botswana is in big trouble
If you look at a map of per capita income, Botswana stands out from the rest of Sub-Saharan Africa. It’s not a rich country, but it’s comfortably middle-income, and does even better than South Africa:
Why is Botswana doing so well? Two reasons: Diamonds, and good policy. Botswana is richly endowed with diamonds, which represent the vast majority of the country’s exports. The country has slowly taken over the international diamond trade, including the company De Beers. And unlike many African countries, Botswana has funneled the money from its mineral exports back into constructive things like education and health. This is made possible by its high level of political stability, which may in turn be a function of its ethnic homogeneity (most Botswanans belong to a group called the Tswana). All in all, Botswana provides a hopeful case that poor resource-exporting countries can escape the dreaded Resource Curse.
But this success could now be under threat. Diamonds are valuable because they’re useful (they look pretty and have many industrial applications too), and because they’re scarce. But artificial diamonds are getting better and better, and soon even the highest-quality diamonds might not be so scarce:
Africa’s trade in natural diamonds [is buckling] under growing pressure from cheaper lab-grown diamonds mass-produced mainly in China and India…Diamond exports, roughly 80% of Botswana’s foreign earnings and a third of government revenue, have tumbled…In September, Botswana’s national statistics agency reported a 43% drop in diamond output in the second quarter, the steepest fall in the country’s modern mining history. The World Bank expects the economy to shrink 3% this year, the second consecutive contraction…
The global rise of synthetic diamonds has been swift…The gems emerged in the 1950s for industrial use. By the 1970s they had reached jewelry quality. Lab-grown stones now sell for up to 80% less than natural diamonds. Once making up just 1% of global sales in 2015, they have surged to nearly 20%…Lab-grown stones now account for most new U.S. engagement rings, he said. Natural diamond prices have fallen roughly 30% since 2022.
This demonstrates a fundamental principle of economic development: In the long term, countries do not become rich from digging up rocks. Humans are ingenious creatures, and we’re always figuring out ways to use resources more efficiently, to substitute common resources for scarce ones, and to use technology to imitate nature. If your country’s prosperity is based on rocks — i.e. on exploiting nature’s bounty in a straightforward, extractive manner — then human ingenuity is working against you.
Botswana’s experience should give pause to the lefty activists who believe that global poverty is due to rich countries exploiting poor countries for their natural resources. Botswana is basically the poster child for the idea that poor post-colonial economies can succeed by taking control of their rocks and selling them at higher prices in global markets. But their success only ever brought them to middle income, and now it’s under dire threat from technological innovation.
The real lesson here is that national wealth comes from ingenuity and hard work, not from sitting on top of rocks.
6. No, prison didn’t replace mental hospitals
Part of a pundit’s job is debunking viral charts. I wrote a two-part series on how not to be fooled by these charts. Today we have another example: the correlation between prison and mental institutions.
This chart comes from a 2013 article in The Economist, which got its data — and its thesis — from a 2011 paper by a law professor named Bernard Harcourt. But in an excellent recent post, Matt Yglesias dug into Harcourt’s papers and showed that the thesis just doesn’t hold up:
Yglesias points out that most of the people going to jail and prison in the 80s and 90s couldn’t have been the same people who were let out of mental hospitals, because a lot of the people in mental hospitals were white women or the elderly:
But (thanks to Xenocrypt for the tip) as Steven Rafael and Michael Stoll pointed out over a decade ago, the population in mental hospitals in 1950 was 87 percent white and 47 percent female. Twenty percent were over the age of 65 in an era when the aggregate population was much younger (only 8 percent of the public were senior citizens)…It is simply not the case that 40 percent of prison inmates are white women or 20 percent of crimes are being committed by senior citizens. We’re just not seeing anything remotely resembling a one-to-one substitution effect here.
Harcourt speculates that mental patients might be raising the crime rate by providing easy victims rather than by actually committing crimes, but Yglesias points out that this doesn’t make sense either, given what we know about victimization statistics.
Yglesias also notes that an earlier Harcourt paper made the much more modest claim that de-institutionalization caused between 4.5% and 14% of the rise in incarceration — in other words, not nearly enough to produce the viral chart.
It’s possible that some people did go from mental institutions to prison when we closed down mental institutions. But this was not the main reason for the rise in incarceration, or even a very big reason. The viral chart is just telling the wrong story.
7. Do Americans hate ugly buildings, or do they just hate buildings?
In the eternal quest to get America to build more housing, it’s important to take politics into account — we have to know how best to avoid activating the ire of local NIMBYs who block housing projects. If you post pictures of new housing online, lots of people inevitably pop up to say the new housing is ugly. And this concern is often raised at various public meetings as well.
But how real are the aesthetic concerns here? Are people just using architecture and style as excuses to block poor people from living near them, or to preserve their quiet streets? A pair of recent papers tried to answer that question by taking surveys, and came up with somewhat conflicting results.
The first paper is by Broockman, Elmendorf, and Kalla. Here’s a thread by Broockman explaining the findings. The authors surveyed people about when they would and wouldn’t support building new housing. Basically, they found that people are fine building more tall apartment buildings in places that are already dense, but oppose it in places that aren’t dense — even if it’s not their own back yard.
Why? The survey respondents said that cities “look nicer” without tall buildings. They also opposed the construction of office buildings, which the authors interpret as meaning that people genuinely care about aesthetics instead of just keeping poor people away. (I’m not so sure; downtown areas may also carry the connotation of being places where lots of poor people hang out.)
Broockman et al. argue that opposition to new buildings can be mitigated by making the new buildings look nice. They find greater support for building nice-looking buildings — the third panel in the pic, instead of the first panel:
I’m a little suspicious of these choices of photos; the first one shows not just an ugly building, but also displays some even taller buildings immediately abutting it, and also some cars on the street that indicate density. The third photo shows trees and gaps between structures, indicating a quiet street and low density. But in any case, the authors think this means that aesthetics matter a lot. They also find that support for building more goes up when they specify that the building would be designed by a star architect.2
Anyway, the second new paper, published just one day earlier, is by Pietrzak and Mendelberg. Here’s a thread by Pietrzak explaining their findings. They also survey a bunch of people about which apartment buildings they would like to see get built.
Two of Pietrzak and Mendelberg’s findings are very similar to Broockman et al. They find that people don’t like tall buildings, but that they support them more if they’re going to be put in neighborhoods that already have tall buildings.
But when it comes to actual building design, their findings were very different from the other paper. Pietrzak and Mendelberg found that there was no preference for traditional brick styles over more modern styles. And whether a tall building has the same style as the buildings around it also has almost no effect; people basically just don’t like tall buildings.
Taken together, these results convince me that Americans believe that tall buildings should be restricted to just a few neighborhoods. That will be an important attitude to overcome, since densifying the inner-ring suburbs will be necessary in order to build significantly more housing in America. But I’m not yet convinced that making buildings look nice makes any real difference to whether Americans will allow them. Further research on that question is needed, and there are probably some natural experiments out there with building code changes along municipal boundaries that could be used to tease out this effect.
8. Is China brain-draining America now?
AI is the world’s most important industry, and China is the source of almost half of the top AI talent in the world:
Domination of the global AI industry is going to largely come down to which country can attract more Chinese talent. Until recently, America was dominating China in this race; most of the best Chinese AI researchers wanted to come and work in the U.S. There has been a lot of (justified) fear that Trump’s policies would reverse this advantage and drive Chinese researchers back to China.
According to a new report by the Carnegie Endowment, brain drain back to China has been very modest so far. Only a small sliver of the Chinese top AI researchers working in America has gone back to their home country since 2019:
The Carnegie Endowment report does highlight a few prominent researchers who have gone back to China, like Yang Zhilin. But so far these are the exception rather than the rule. And their return isn’t even necessarily due to Trump — China is pouring money into the AI industry, and companies like DeepSeek have emerged as real competitors to American industry leaders.
The Carnegie report explains that the bigger worry for the U.S. isn’t brain-drain — it’s that Chinese talent won’t even come here in the first place:
Although the United States has managed to retain a large portion of Chinese AI researchers over the past six years, there are signs that it has lost some of its ability to attract new arrivals from China—a potentially ominous trend given China’s share of global AI talent…
By [2022], Chinese-origin researchers made up nearly half of all the authors sampled, and Chinese institutions had more than doubled their share to 28 percent [since 2019]. That was still well short of the U.S. share of 42 percent, but it demonstrated rapid catch-up by China in producing many of the year’s best AI research papers.
Everyone seems to be sleeping on just how much China dominates the production of global top AI talent. If other countries can’t improve their own talent pipelines, their AI industries will continue to be dependent on luring Chinese people out of China.
Gelman also uncritically repeats Wainer et al.’s claim that Mississippi ranked 50th in the nation for 4th-grade math. In fact, as Piper points out, Mississippi actually ranked 16th. Wainer et al. simply made a mistake.
Substack hosts a semi-regular series of live debate events, where I had the pleasure of participating in a live debate with fellow substacker on the topic of “Should robots take our jobs?”. I officially won the debate, but in the end, I think Brian and I agreed on more than we disagreed on.
We agreed that it’s unlikely that robots actually do take all our jobs; in other words, we’re arguing about a fairly sci-fi future instead of a likely scenario. And our basic stance is that society needs to develop institutions to make sure that the wealth from automation is widely distributed throughout society. In the Industrial Age, those institutions were things like welfare states, taxes, unions, labor regulations, minimum wage laws, and so on. It’s not clear what new redistributive institutions would be necessary in an age of pervasive automation — Sovereign wealth funds? UBI? Resources set aside for human consumption? — but it seems likely that we would need some new ones.
Where Brian and I disagree is on the likelihood of these institutions being developed, and on the pain and suffering that will be required in order to build them. Brian thinks that as long as rich guys like Sam Altman are in charge of the development of AI, it will be hard to change society to “redistribute the robots”. I’m far more optimistic; I don’t think the Sam Altmans of the world will ultimately have that much power over our institutions.
It’s also a question of whether you take the long view or the short view in terms of how this all shakes out. In the case of the Industrial Revolution, it took centuries of social and political struggle to wrestle the new productive system into something egalitarian, and along the way there were some very horrific failures such as communism. Whole lifetimes and whole generations were swallowed up by the struggle to tame industrial technology.
But in the end, we succeeded. Our societies are immeasurably better than if we had simply shied away from inventing power looms, or machine tools, or harvesters, or any of the other labor-saving automation that took most of our ancestors’ jobs. Stopping a new technological revolution in its tracks, or significantly slowing it down, is very hard. In general, the only way out is through — the best world comes not from resisting new technology, but from accelerating the development of sociopolitical institutions that make sure new technology’s benefits are widely shared.
I think the audience at the debate agreed with my more optimistic, long-term perspective.
Anyway, big thanks to the Substack Events Team for producing the event, and especially to for getting the event together in the first place. It was very fun, and I’m looking forward to the next one!
I’ve actually already written a number of posts about the possibility of an AI bubble and bust. Back in August, I wondered if the financing of data centers with private credit could cause a financial crisis if there was a bust. I followed that up with a post about profitability, and suggested that the AI industry might be a lot more competitive than people expect. In October, I wrote about how AI is propping up the U.S. economy.
But I feel like I need to write another post, because almost all of the discussion I see about an AI bubble seems to leave out one crucial scenario.
Since I wrote those posts, popular belief that there’s an AI bubble and impending bust has only grown. A lot of prominent people in the industry are talking about it:
“Some parts of AI are probably in a bubble,” Google DeepMind CEO Demis Hassabis told Axios’ Mike Allen at the [Axios] AI+ Summit on Dec. 4. But, he added, “It’s not a binary.”…“I, more than anyone, believe that AI is the most transformative technology ever, so I think in the fullness of time, this is all going to be more than justified,” Hassabis said…“I think it would be a mistake to dismiss [AI] as snake oil,” OpenAI Chairman and Sierra co-founder Bret Taylor said at the AI+ Summit…Taylor acknowledged that there “probably is a bubble,” but said businesses, ideas and technologies endure even after bubbles pop. “There’s going to be a handful of companies that are truly generational,” Taylor said.
Every company would be affected if the AI bubble were to burst, the head of Google’s parent firm Alphabet has told the BBC…Speaking exclusively to BBC News, Sundar Pichai said while the growth of artificial intelligence (AI) investment had been an “extraordinary moment”, there was some “irrationality” in the current AI boom.
Almost everyone I read is basically talking about two scenarios for an AI bust. I call these the Virtual Reality Scenario and the Railroad Scenario. I’ll go over these, and then talk about the third scenario
The Virtual Reality Scenario
What I call the Virtual Reality Scenario is if AI, in its current form, turns out to just not be a very useful technology at all — or at least, not nearly useful enough to justify the amount of capital expenditures. This might happen because AI hallucinates too much, or because progress in AI comes to a halt. Bloomberg reports:
The data center spending spree is overshadowed by persistent skepticism about the payoff from AI technology…[R]esearchers at the Massachusetts Institute of Technology found that 95% of organizations saw zero return on their investment in AI initiatives…More recently, researchers at Harvard and Stanford [found that e]mployees are using AI to create “workslop,” which the researchers define as “AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.”…
AI developers have also been confronting a different challenge. OpenAI, Anthropic and others have for years bet on the so-called scaling laws — the idea that more computing power, data and larger models will inevitably pave the way for greater leaps in the power of AI…Over the past year, however, these developers have experienced diminishing returns from their costly efforts to build more advanced AI.
This is basically what happened to VR technology. Meta spent $77 billion on developing the virtual reality “Metaverse”, but outside of gaming and some niche entertainment applications, nobody really wanted VR for anything, no matter how good the headsets got. Meta is now throwing in the towel and pivoting away from the Metaverse.1
But I just don’t think this is going to happen to AI. Very few people use VR, even a decade after it started getting hyped. But AI is being adopted more rapidly than any technology in history. As far back as a year ago, 40% of people were already using AI at work:
Humans just know when a technology works. If AI weren’t useful, we’d see people trying it for a while and then setting it aside. But we don’t see that. Despite hallucinations and the other limitations of the technology, most people are finding some reason to keep using AI after they try it.
Sorry, haters. This tech is for real.
As for progress hitting a wall, I just don’t think that’s an important question anymore. The consensus in the industry2 seems to be that scaling up in terms of training AI models on more data has hit a wall, but that our ability to improve AI’s capabilities via inference scaling (basically, having it “think harder” before answering) is still going, and improvements in reinforcement learning and other algorithmic techniques are still coming.
But I don’t think this is the right question to be asking, because “better chatbots” are only one of many ways that AI can create value. The world of AI applications, including “agents” (AI that does stuff on its own) is still in its infancy. Andrej Karpathy is good at talking about this; I recommend his interview with Dwarkesh Patel.
Essentially, we haven’t even begun to build AI yet. Anthropic sort of has — it’s focused on AI business applications, and it’s making some money doing this. But most of the actual technology we’re going to create with AI is still in the future. And there’s a chance that AI as it exists today will be able to provide the foundation for a whole bunch of incredibly useful applications, even without any continued improvement in capabilities. The typical pattern is for most of the useful (and lucrative) applications to manifest decades after a new general-purpose technology is introduced.
The Railroad Scenario
Which brings us to the second scenario: the Railroad Scenario. Railroads were even more economically useful and lucrative than their biggest boosters in the 1860s imagined. But there was still a huge bust in 1873, because those economic benefits didn’t show up before railroad companies’ debt came due.
The railroad buildout in the 1800s was, in percentage terms, the greatest single feat of capital expenditure in U.S. history, dwarfing even what the AI industry is spending on data centers right now. In 1873, a bunch of railroad-related loans went bust, causing a banking crisis that threw the economy into a decade-long depression.
And yet despite the carnage in the railroad industry, the number of miles of railroad in the U.S. never stopped increasing — it just slowed down briefly and picked back up again!…In other words, the great railroad bust did not happen because America built too many railroads. America didn’t build too many railroads! What happened was that America financed its railroads faster than they could capture value…
There’s basically no technological limit on how many…loans [the financial system] can disburse in a short period of time…And yet there is a limit on how fast businesses can create real value. In order for a railroad to pay off, you need to build it, and then you need to find people to pay you to ship things on it. That takes time, especially because the full economic value of the railroad doesn’t manifest until new cities, new industries, and new supply chains that are enabled by the railroads get created.
Just to give one example, the famous Sears Catalog — which allowed people all across America to order products and have them delivered by railroad — eventually revolutionized American retail. But it didn’t even start to do that until 1888 — fifteen years after the big railroad crash of 1873…
Why is this important for AI? Because even if AI creates all the value its biggest boosters say it’ll do — supercharges growth, enables the automation of most kinds of production, and so on — it might not do it fast enough for the data center “hyperscalers” to pay back everything they borrowed. In that case, there will be a wave of defaults on bonds and loans.
We don’t know how likely this scenario is, because we don’t know how fast AI value creation will increase. But we can get some basic idea of the risk of this scenario by looking at the financing side.
If the companies building and operating the data centers (the main cost of AI) are spending less than they make, then everything is basically safe. Suppose you have a company that makes $50 billion in profit every year, and you spend $40 billion every year on data centers. Even if AI suffers a catastrophic crash and all your money is wasted, you’re safe; you just take a hit to your profit margins for a year, your stock price goes down, and then you just move on. This is true even if you borrowed the money to build the data centers; if things go bad you can afford to pay off the loans.
If you’re spending $70 billion a year things get dicey; you might have to take a couple of years of losses to pay back the loans if there’s a bust. And there’s some level of borrowing and spending where you’re actually in danger of bankruptcy. There’s no hard and fast rule for when a certain amount of spending becomes dangerous; it’s just sort of a sliding scale of worry.
Right now, much of the AI buildout is being done by big tech companies like Google, Microsoft, Amazon and Meta that make lots and lots of profit.3 Until recently, these “hyperscalers” have been making enough cash to cover their AI spending. But spending is rising, so that may not be true for much longer. If spending keeps increasing, some companies, like Amazon, may start to have to borrow against future cash flow soon. And Meta, which doesn’t have its own cloud business, and thus has to pay other companies to do its AI stuff, may be in greater danger.
Meanwhile there are a bunch of other companies that are investing a ton of money in AI that don’t make enough profit to fund it out of their own pockets, but which have borrowed a bunch of money to invest in AI. These include the big model-making companies — OpenAI, Anthropic, and xAI. They also include some cloud providers like CoreWeave that aren’t attached to a big profitable business. And they include various construction companies and service providers.
If AI takes 10 or more years to generate enough value to pay back all these debts, many of these companies could go bankrupt. Whatever financial institutions they’ve borrowed money from — private credit firms, banks, etc. — may fail or have to pull back significantly when their loans suddenly go bad. And that could touch off a financial crisis, even if AI continues to advance and companies continue to build data centers. That would be like what happened in 1873 with the railroads. AI itself would be fine, but the economy, the financial system, and a bunch of specific companies could be hurt.
This scenario seems at least fairly likely, given that this is basically what happened with both the railroads and the telecoms in past industrial booms. Currently, many observers are highly skeptical that AI companies will be able to earn enough revenue to pay back their debts by 2030, even under optimistic assumptions. As I said, the typical pattern is for the economy to take a long time to figure out how to use each new general-purpose technology, and the financial system may not be able to wait around.
But there’s also a third scenario, which relatively few people seem to be paying attention to. Even if AI works and manages to create value very quickly, that value may not be captured by the AI companies themselves. AI itself may turn out to be a commoditized, low-margin business, more like solar power…or airlines.
The map above is a depiction of The Deluge, a historical event in which the Polish-Lithuanian Commonwealth — which had been a major European power — was defeated and destroyed under the combined assaults of Russia and Sweden in the 1600s. After having its power broken, Poland was carved up in the 1700s and subjugated by Russia, Prussia, and Austria. It took more than two centuries, until the fall of communism in 1991, for Poland to reemerge as a strong, truly independent country.
The Deluge shows that power and independence are not permanent. If you are surrounded by hostile powers, and if you don’t have the ability to guard yourself against those powers, no amount of historical greatness can save you from being subjugated. This is an important lesson for Europeans to remember right now, as they find their region under siege from Russia, China, and the United States all at once.
The United States no longer cares about the European project
Why would America care about Europe at all? For most of our history, we didn’t. In the 19th century, the U.S. viewed European countries as dangerous rivals. In the early 20th century, Americans prided themselves on not getting involved in European affairs, and were incensed at their government for dragging them into World War 1. Only after World War 2 did Americans start caring about Europe, and we did so for three reasons:
West Europe was a bulwark against Soviet communism.
Europe was a key trading partner.
Many Americans came to value their ancestral ties to Europe.
The first of these reasons vanished in 1991. Europe is still a bulwark against Russia, but Americans no longer feel threatened by Russia. Russian power is far less than what it once was, and Russia’s rightist ideology does not threaten the rightists who now rule America.
As for communism, many (most?) Americans now believe that European countries are socialist. When American conservatives ask where in the world socialism has succeeded, American progressives will always reply “Europe” or “Scandinavia”. Whether Europe or Scandinavia is actually socialist is irrelevant; Americans have come to see it that way.
Europe is still an important trading partner. But Trump and the other people now in charge of the U.S. do not understand trade at all. They think about trade entirely in terms of the net trade balance, rather than in terms of total U.S. exports. Trump & co. don’t care that America sells $650 billion a year to Europe; the fact that Europe sells $800 billion a year to America means that Trump & co. think America is “losing” and would benefit from a cutoff of trade.
Remember that the U.S. is an unusually closed-off, self-sufficient economy, so Americans in general don’t think too hard about trade or try to understand why it’s valuable. Also, the people running now the country are especially ignorant about economic matters.
As for civilizational ties, this is the reason Trump and the MAGA movement have turned so strongly against Europe. The American right values Europe because they think of it as a White Christian homeland — the source and font of Western civilization. Here’s a post I wrote about that earlier this year:
I wrote:
in the American mind, Europe stood across the sea as a place of timeless homogeneity, where the native white population had always been and would always remain…In the mind of many Americans, Europe thus stood as both a refuge and a reservoir. America itself was a rough, contested frontier, but Europe would always be white and Christian. If you ever felt the need to live around a bunch of white people of Christian heritage, you could always go “back”, but for most that wasn’t necessary — just knowing that the Old World was somewhere out there was enough.5
I think Europeans may underestimate how much this perception motivated America’s participation in the Transatlantic Alliance during the Cold War…[T]o conservative Americans in the 20th century — the type of people who joined the John Birch Society — the Cold War was about preserving Christendom from the threat of godless communism.
Anyway, in the 2010s, it dawned on those Americans that this hallowed image of Europe was no longer accurate. With their working population dwindling, European countries took in millions of Muslim refugees and other immigrants from the Middle East and Central and South Asia — many of whom didn’t assimilate nearly as well as their peers in the U.S. You’d hear people say things like “Paris isn’t Paris anymore.”6…At the same time, Europe had long since abandoned its traditional Christian values…
To Americans who valued the idea of America and Europe as part of a single Western civilization, this realization was catastrophic. Suddenly European countries — and the Anglosphere countries of Canada, Australia, and New Zealand — felt like they had left the club…
America’s rightists…want to know that someone, somewhere, is out there preserving an indigenous homeland for their identity groups. And that “someone” has to be Europe and the Anglosphere.
This isn’t a new attitude, either. Remember that in order to persuade a reluctant America to join World War 1, the U.S. government had to depict Germany as an ape abducting a white woman!
If you understand this, then nothing in America’s new National Security Strategy is mysterious, surprising, or confusing. Here’s how War on the Rocks summarizes the Trump administration’s attitude toward Europe:
[I]mmigration is elevated to the central national security problem. The text declares, bluntly, that “the era of mass migration must end,” and that “border security is the primary element of national security.” It frames mass migration as a driver of crime, social breakdown, and economic distortion, and calls for a world where sovereign states cooperate to “stop rather than facilitate destabilizing population flows” and tightly control whom they admit…
[P]rotecting American culture, “spiritual health,” and “traditional families” are framed as core national security requirements…The document insists that “restoration and reinvigoration of American spiritual and cultural health” are prerequisites for long-term security and links this to an America that “cherishes its past glories and its heroes” and is sustained by “growing numbers of strong, traditional families” raising “healthy children.” America is thus cast as defender of so-called traditional values, while Europe lacks “civilizational self-confidence and Western identity.”…
[T]he strategy elevates the culture wars into a governing logic for national security, and it does so through rhetoric that treats ideological and cultural disputes as matters of strategic consequence…This is clearest in the European section…The text…speculates about demographic and cultural shifts in Europe as a way to question whether future governments will share American views of their alliances. The strategy [implies] that cultural alignment is essential to strategic partnership.
The American right sees the “mad brute” in the ape cartoon as the dark-skinned Muslim immigrants who have entered Europe in large numbers in recent years. And they see themselves as needing to save the woman — representing their view of Europe as the traditional font of White Christian civilization — from that mad brute.
This tweet by Elon Musk pretty much sums up the American right’s attitude toward Europe:
This is why no amount of European shaming or moral persuasion can have any effect on the Trump administration — or on any Republican administration in the decades to come. This kind of appeal to friendship is totally useless:
The American right — i.e., the people now in charge of the country — do not care intrinsically about democracy, or about allyship, or about NATO, or about the European project. They care about “Western Civilization”. Unless Europe expels Muslim immigrants en masse and starts talking about its Christian heritage, the Republican Party is unlikely to lift a hand to help Europe with any of its problems. Democrats will want to help Europe, but they will only be in power intermittently, and helping Europe will not be high on their priority list.1
Thus, America is not riding to the rescue this time, or for the foreseeable future. I wish things were different, but my wishes count for nothing; this is the reality with which the Europeans must now deal.
Russia and China together are the real menace to Europe
Europeans do not need me to tell them that Putin’s Russia threatens not just Ukraine, but all of Europe. They are well aware of this fact. Russia now regularly flies its drones into Europe, and is probably behind a wave of sabotage attacks on European infrastructure.
How can Russia, a country of just 144 million people and $7 trillion in GDP (PPP), hope to overcome Europe, which has 520 million people and $33 trillion in GDP (including the UK), especially after Russia has expended so many of its young men and materiel in its war with Ukraine already? There are three answers here. The first is gray-zone warfare, including sabotage and political influence campaigns. But that’s only the beginning.
Russia’s second method for fighting Europe is what I call a “Ponzi empire” strategy. Russia has enslaved vast numbers of Ukrainians from the occupied regions of Ukraine to fight against the rest of their country. If Russia conquers the rest of Ukraine, it will similarly enslave the rest of the country’s population, and send them to fight against Poland, the Baltics, and Moldova. If they then defeat Poland, they will enslave the Poles and send them to fight against the next European target, and so on.
This is a very traditional Russian strategy. Enslaved Ukrainians were used to attack Poland in 1939. Enslaved Poles were forced to fight Russia’s wars in the days of the old Tsarist empire, and would have been forced to do so again as part of the Warsaw Pact. Just like zombies turn humans against their own, each slice of Europe that Russia can chop off ends up being turned against the rest.2
Russia’s final strategy for fighting Europe is to rely on Chinese assistance. Russia’s own industrial base is very weak, and relied heavily on imported European parts and machinery that has now been partially cut off. But Chinese tech has largely plugged that hole, as the Carnegie Endowment reports:
Since mid-2025, Chinese components have been detected in Russian drones and missiles, often shipped via front companies disguised as suppliers of industrial cooling equipment…Chinese machinery, including precision optics, lasers, and dual-use machine tools, now dominates Russia’s defense-related manufacturing. In August 2025 alone, China exported a record 328,000 miles of fiber-optic cable and nearly $50 million worth of lithium-ion batteries to Russia, reinforcing its role as the Kremlin’s primary wartime supplier of dual-use materials. Chinese engineers working at Russian drone facilities are adapting civilian quadcopters, such as the Autel Max 4T, for combat use.
China is a far bigger manufacturer than Europe, and can pour essentially infinite war production into Russia if it wants to. And China is now assisting Russia’s gray-zone warfare against Europe:
Since 2024, Chinese ships have been involved in incidents of targeting subsea infrastructure, particularly cutting subsea cables in the Baltic Sea…The country increasingly deploys ambitious espionage and cyber attacks against government networks and critical infrastructure across Europe. These attacks seem to overlap with—or even be actively coordinated with—Russia’s espionage and influence operations across Europe…Increasingly, Russia and China also cooperate in disinformation operations: Chinese campaigns such as “Spamouflage” are amplified by Russian media outlets and diplomatic channels. Both countries employ what look to be synchronized narratives accusing the West of being responsible for the war in Ukraine.
China even provides the Russians with battlefield intelligence, helping them strike and destroy Ukrainian targets in real time. In sum, China is supporting Russia’s war against Ukraine, and will likely support Russia in any further wars it undertakes against the rest of Europe.
With Chinese technology and production, and slave soldiers from East Europe, and with America withdrawing from the Transatlantic Alliance, Russia could conceivably overmatch Europe.
But that’s not the only threat that China poses. On the economic front, China’s new economic strategy — a combination of shutting out European products, sending out a massive wave of subsidized exports, and putting export controls on rare earths — threatens to forcibly deindustrialize Europe. Here’s what The Economist, normally a staunch defender of free trade, recently wrote:
China is not just dumping exports and subsidising its companies, it is also out-competing and out-innovating big European industries, including carmaking. Last year Germany’s trade deficit with China stood at €66bn ($76bn); this year it could widen to over €85bn, around 2% of GDP. Alarmingly, China is exploiting Europe’s dependence, weaponising embargoes or the threat of them in chips and rare earths.
Germany, traditionally Europe’s strongest manufacturing and exporting nation, is already the hardest hit:
China, many European manufacturers have concluded, is threatening to put them out of business, by both fair means and foul…The wails are loudest in Germany, which is Europe’s biggest exporter to China and its biggest investor in it by far…For the Mittelstand, the small manufacturers that constitute a big slice of German industry, China used to be a source not of angst but of profit. Their precision-engineered machine tools were an exquisite fit for its rapid industrialisation. Chinese consumers raced to buy German cars…
Times have changed…Once-stellar growth inside China has, for many foreign firms, slowed to a crawl as competition with local rivals intensifies. In addition, Germany’s previously small trade deficit with China has ballooned…Last year it reached €66bn ($76bn), or around 1.5% of GDP, driven by a collapse in German exports to China and a rush of imports, notably of cars, chemicals and machinery—hitherto German specialities.
Germany’s trade deficit with China this year is expected to surge again, to around €87bn…German cars command only 17% of the Chinese market, down from a peak of 27% in 2020…Worse, Chinese competition also jeopardises sales in other markets. China’s net exports of cars have risen from zero in 2020 to 5m units last year. Germany’s have halved over the same period, to 1.2m units…Such figures have triggered fears in Germany of a wave of deindustrialisation.
This is all on top of the existing headwinds facing European manufacturing — the energy crisis from the cutoff of Russian gas and self-inflicted “green” policies, Trump’s tariffs, and so on.
So Europe finds itself in an extraordinary perilous position right now. Its main protector has suddenly withdrawn. It has a ravenous, brutal empire attacking its borders, supported by the world’s most powerful nation. Its main export markets are shriveling, and its manufacturing industries are under dire threat from waves of subsidized foreign competition. What can it do to fight back?
How Europe can resist the siege
The most important thing Europeans need is to panic. Europe is facing its own Deluge — a sudden pincer movement by hostile great powers that threatens to reduce it to a collection of small vassal states. This is a true crisis, and it will not be solved by social media rhetoric, or by brave declarations by EU leaders. It cannot be regulated away by eurocrats in Brussels. It will require bold policies that change Europe’s economic, political, and social models. Only a strong sense of urgency and purpose can motivate Europe to do what needs to be done.
What needs to be done? One important step is for Europe to act more as a single whole than as a collection of small countries. In the military realm, this means coordinating European militaries and defense industries much more. Matthew C. Klein writes:
From a properly European pespective, the security interests of each country should be shared across all countries, just as, for example, most Americans in Michigan or Maine would view an attack on California or Florida as an attack on them…The first step is to give the Ukrainians, who are already fighting the Russians, as much material and financial support as they need. From the perspective of European security, French, German, and British weapons are far more valuable in Ukraine than in their home countries. If the Ukrainians were subjugated, defending the rest of Europe would become much harder, with the effective EU-Russia border lengthening dramatically…
Europe’s national militaries have had a tendency to favor their home country’s producers, with the result that the continent is filled with subscale defense companies that are often slow and unproductive. Common defense procurement for a continental army should lead to higher output and lower costs—a few large companies handling large orders should have better unit economics than hundreds of artisanal manufacturers—but it would require Europe’s national defense elites to change their perspective. Philipp Hildebrand, Hélène Rey, and Moritz Schularick recently published a useful proposal for how to make this work.
And economically, Europeans can partially compensate for the loss of Chinese (and American) export markets by selling more to each other. The Economist writes:
A second task is for European countries to make better use of the power they have, by integrating their economies…By failing to integrate, the EU is leaving a vast sum of money on the table. A single market that was designed for goods is failing to help economies dominated by services.
And in his famous report on European competitiveness, Mario Draghi wrote:
We have also left our Single Market fragmented for decades, which has a cascading effect on our competitiveness. It drives high-growth companies overseas, in turn reducing the pool of projects to be financed and hindering the development of Europe’s capital markets…The EU’s new industrial strategy rests on a series of building blocks, the first of which is full implementation of the Single Market. The Single Market is critical for all aspects of the strategy: for enabling scale for young, innovative companies and large industrials that compete on global markets; for creating a deep and diversified common energy market, an integrated multimodal transport market and strong demand for decarbonisation solutions; for negotiating preferential trade deals and building more resilient supply chains; for mobilising greater volumes of private finance; and as a result, for unlocking higher domestic demand and investment. Remaining trade frictions in the EU mean that Europe is leaving around 10% of potential GDP on the table, according to one estimate.
And ideally, Europe should form a fiscal union — the EU itself should be able to borrow and spend, not just the member countries. As Klein writes, this needs to be accompanied by a greater tolerance for fiscal deficits — after all, countries borrow in emergencies.
In other words, Europe’s first step in resisting its siege is to act more like a country and less like a zone. It would also help to find some way to bring the UK back into the fold, especially because polls consistently find that British people regret Brexit.
Europe’s other top priority is to provide for the common defense. That means spending more money on the military, of course, and it also means greatly increasing the size of Europe’s nuclear deterrent. But it also means building a defense industrial base capable of resisting a China-backed Russia.
Europe’s current defense-industrial base was built for the Cold War, when battles were decided by heavy vehicles like tanks and ships and planes. Those are still somewhat important, but drones have risen very quickly to dominate the modern battlefield. Right now, drone manufacturing, as well as almost the entire supply chain for battery-powered drones, is overwhelmingly concentrated in China.
Europe needs to be able to build not just drones, but every single thing that goes into making a drone — batteries, motors, various types of computer chips, and so on. European industrial policy should therefore focus on onshoring these industries. In other words, Europe needs to master the entire Electric Tech Stack. (This will also help Europe get back in the EV race.) And it needs to master the AI software — computer vision, swarming tech, and so on — that will soon be needed in order to make drones a truly modern force.
The question of the proper policy instrument to accomplish this goal — tariffs, subsidies, fiscal borrowing, regulatory changes, and so on — is irrelevant. All of these policies should be done as necessary, and it’s better to do too much than too little. Policy procedure needs to be subordinated to the overriding goal of making Europe capable of defending itself. In fact, every European institution needs to be reformed and reverse-engineered in order to enable this.
Europe is also going to have to change its political mindset. Lavish pensions and other elements of Europe’s social model are going to have to be temporarily curbed to help give Europe the fiscal space and physical resources to fight off its enemies. All nuclear plants need to be restarted, and Europe should build more nuclear, ignoring “green” parties and environmental activists who irrationally hate nuclear power. Europe needs to reform its land-use regulation to require greater construction of solar and wind power. And Europe is going to have to back off of its aggressive regulation of AI software, in order to produce cutting-edge autonomous weaponry.
Finally, Europe needs to look for friends and allies — and export markets — other than America. India is an obvious choice. Although India is friendly with Russia, the country would undoubtedly welcome Germany’s help industrializing — and this would allow German companies to sell machines to India, as they once did to China. The EU should open its markets to Indian goods in exchange for Indians doing the same, recognizing that trade balances are less important than total export demand. Japan, South Korea, and other big developing countries like Indonesia, Vietnam, and Brazil are other good potential trading partners.
If Europe manages to unify more and to build up its military power, it will increase the number of great powers in the world by one. A planet with a strong Europe, America, China, Russia, and India is a better planet than one where only the last four of those are strong. If Europe shows it can act with unity and purpose, and that it has military power to be reckoned with, America and China — both countries whose leaders tend to respect raw power — may lose their disdain for the region, and return to a more diplomatic, conciliatory posture.
Ultimately, European weakness and division are the reasons the region is getting bullied by so many other powers. Reversing that weakness and division would make the bullies go away. But Europe’s people, and especially Europe’s elites, have to want it.
And of course if Europe does expel the Muslim immigrants and start talking up its Christian heritage, as the MAGA folks want, Democrats will conclude that Europe is fascist and be reluctant to help it out when they get back in power. Essentially, Europe is finding itself caught in America’s internal culture wars, and there’s no good way out; the only solution is to realize that the U.S. will not be a reliable partner for decades to come.
Would Russia actually try to conquer and rule all of Europe directly, as the Nazis tried to do? Unlikely. But would it try to dominate all of Europe the way the USSR dominated the Warsaw Pact? Yes, definitely. And this sort of domination would be very bad for Europeans, as the Poles could tell you.
In the 2010s, a bunch of right-wing types suddenly became big fans of Martin Luther King Jr.’s views on race. If you saw someone on Twitter quote MLK’s nostrum that people should “not be judged by the color of their skin but by the content of their character”, it was almost certainly someone on the right — quite a change from the type of person who probably would have cited King’s words half a century earlier. This is from an Associated Press story back in 2013:
King’s quote has become a staple of conservative belief that “judged by the color of their skin” includes things such as unique appeals to certain voter groups, reserving government contracts for Hispanic-owned businesses, seeking more non-white corporate executives, or admitting black students to college with lower test scores.
Many progressives railed against the idea of a colorblind society, arguing that statistical disparities between racial groups — income gaps, wealth gaps, incarceration gaps, and so on — couldn’t be remedied without writing race into official policy and becoming much more race-conscious in our daily lives.
In that kind of environment, it’s understandable that lots of people on the right would turn to individualist principles like the ones espoused by MLK in his famous speech. Asking to be judged by the content of your character is a reasonable defense against people who are trying to judge you based on your membership in a racial group.
[I]t would be a shame if this single act of betrayal became the excuse for deporting all Afghan refugees in the U.S…Tens of thousands are building new lives here in peace and are contributing to their communities. They shouldn’t be blamed for the violent act of one man.
Stephen Miller, Trump’s powerful Homeland Security Advisor, responded with a dismissal of individualism and an indictment of Afghans as a group:
This is the great lie of mass migration. You are not just importing individuals. You are importing societies. No magic transformation occurs when failed states cross borders. At scale, migrants and their descendants recreate the conditions, and terrors, of their broken homelands.
Federal prosecutors charged dozens of people with felonies, accusing them of stealing hundreds of millions of dollars from a government program meant to keep children fed during the Covid-19 pandemic…At first, many in the state saw the case as a one-off abuse…But…Over the last five years, law enforcement officials say, fraud took root in pockets of Minnesota’s Somali diaspora as scores of individuals made small fortunes by setting up companies that billed state agencies for millions of dollars’ worth of social services that were never provided…Federal prosecutors say…more than $1 billion in taxpayers’ money has been stolen[.]
In the wake of those revelations, Trump condemned Somalis as a group:
Here are Trump’s exact words:
I don’t want em in our country. Their country’s no good for a reason. Their country stinks…I could say that about other countries too…We don’t want em…We have to rebuild our country. You know, our country’s at a tipping point. We could go bad…We’re going to go the wrong way if we keep taking garbage into our country. Ilhan Omar is garbage. Her friends are garbage…And from where they came from, they got nothing…When they come from Hell, and they complain, and do nothing but bitch, we don’t want em in our country. Let em go back to where they came from, and fix it.
Here you see the very same idea that Stephen Miller expressed. Trump and Miller both judge people by their ethnic group, and they judge those ethnic groups by the condition of their ancestral country. Somalia is a bad place, therefore Somalis are bad, therefore if you’re a Somali you’re bad and you shouldn’t be allowed into America. Afghanistan is a bad place, therefore Afghans are bad, therefore if you’re an Afghan you’re bad and you shouldn’t be allowed into America.
In fact, this idea was very popular a century ago, when America enacted harsh restrictions on immigration. Restrictionists argued that immigrants from South and East Europe were undesirable, because South and East Europe were relatively underdeveloped places. For example, here’s what Francis Walker, the president of MIT and a staunch opponent of immigration, wrote in The Atlantic in 1896:
Only a short time ago, the immigrants from southern Italy, Hungary, Austria, and Russia together made up hardly more than one per cent of our immigration. To-day the proportion has risen to something like forty per cent…The entrance into our political, social, and industrial life of such vast masses of peasantry, degraded below our utmost conceptions, is a matter which no intelligent patriot can look upon without the gravest apprehension and alarm. These people have no history behind them which is of a nature to give encouragement…They are beaten men from beaten races; representing the worst failures in the struggle for existence…They have none of the ideas and aptitudes which fit men to take up readily and easily the problem of self-care and self-government. [emphasis mine]
This is a form of racial collectivism. It’s judging people by ethnic, racial, and national groups instead of as individuals. In his landmark 1955 book Strangers in the Land, which chronicles the anti-immigration movement of the late 19th and early 20th centuries, historian John Higham labeled this attitude “racism”. Today, of course, we can’t use that word, since it has been repurposed to mean so many other things. But the word feels like a perfect fit — it’s an ideology (an “ism”) that holds that people are to be judged according to the collective accomplishments of their race.
When I see people on the right spouting this sort of rhetoric, I think: What happened to MLK? What happened to judging people based on the content of their character? What happened to the colorblind society? What happened between 2018 and now that makes collective judgment of racial groups suddenly ok?
The answer, of course, is “The right got the upper hand in American politics.” It turns out that individualism is a bit like free speech — a principle that lots of people tend to support when their tribe is losing, only to abandon it as soon as they’re back on top. A lot of people really do believe in individualism, of course, especially in America. But a lot of others just use it as a cynical shield when they’re on the defensive. And we’re finding out that most of the MAGA movement was always the latter type.
MAGA’s overriding goal is immigration restriction. They care about this much more than any other policy issue — more than inflation, more than trade, more than crime, more than anything. And the reason they want immigration restriction, I believe, is because they think that Somalis and Afghans and Haitians and so on are going to make America more like those countries. When Trump and Miller talk about this, I think they’re being completely honest. And after Trump is gone, I think this idea will be at the core of the new right-wing ideology that will sustain the MAGA movement. Racial collectivism is absolutely at the core of their worldview.
But MAGA has a big problem: While that worldview has some appeal to Americans, overall they aren’t on board. Every poll we have shows pro-immigration sentiment on the rise again, after a dip during the Biden years:
A lot of Americans are also in favor of individualism — that is, of treating people based on their individual traits rather than what group they belong to. Americans of most races supported the recent Supreme Court decision banning racial preferences in university admissions; even black Americans were about evenly split. And while Americans disagree about lots of racial issues, they tend to overwhelmingly say they support things like equal opportunity regardless of race.
And although there are differences in American attitudes toward immigrants from different regions of the world, the differences aren’t huge, and they don’t perfectly line up with how developed the regions are. For example, here’s a 2015 poll by Pew, finding that immigration from Africa is viewed more favorably than immigration from the much more developed regions of Latin America and the Middle East:
So although some Americans are probably evaluating immigrants based on their racial group and on the condition of their source country, like Trump and Miller are, Americans in general probably don’t think this way. They get mad at illegal immigration, and at the disorderly quasi-legal immigration that Biden tolerated — but illegal entry is an individual action, not a group trait.
Which makes sense. The U.S. immigration system is highly selective; Lazear (2017) shows that selectivity accounts for a very large fraction of the average educational attainment of different immigrant groups in America.
As Matt Yglesias points out, nowhere is this more evident than with Indian immigrants. The country of India is still poor; despite solid recent growth, its GDP per capita is lower than that of El Salvador or Guatemala. Infrastructure has improved a lot but is still subpar, and the country has pockets of startling poverty. By the racial-collectivist logic of Miller and Trump, or of the restrictionists of a century ago, Indian immigrants should be turning America into a third-world country.
Nor has Indian immigration turned anywhere in America into a version of India. Fremont, California is probably the city over 100,000 population with the greatest percentage of Indians — about 29%. And yet Fremont is one of the cleanest, nicest, richest, safest towns in the whole country, with a murder rate so low that many European countries would envy it, and arguably the best public schools in the country. A recent survey identified Fremont as the happiest city in America.
A big part of this, of course, is because immigration from India is so selective. India is the world’s most populous country; it’s not too hard to grab a few million smart people from a country that big. But this isn’t the only reason. American institutions are also important.
As another example, take El Paso. The overwhelming majority of people in El Paso are of Mexican descent. Mexican immigration is among the least selective, because Mexico is so close to America and there was so much illegal immigration in the past. And yet despite being filled with ethnic Mexicans, El Paso looks absolutely nothing like Juarez, the Mexican city that sits right next to it on the opposite side of the border. El Paso’s murder rate is 3.8, very low for an American city, while Juarez is one of the most violent, chaotic cities on planet Earth.
Mexicans didn’t turn El Paso into Mexico, and the reason is American institutions. America’s economy offers El Paso’s residents the chance to get ahead without joining drug gangs. American culture is a more positive-sum, less violent culture than Mexico’s. And the U.S. Military has a big presence in El Paso, because Fort Bliss is there. Even without selectivity, institutions matter a lot.
So Stephen Miller is just flat-out wrong. Immigrants do not recreate the conditions of their homelands in America. Yes, there is some amount of carryover, including some negative influences like the old Sicilian mafia, or modern gangs like MS-13. But the differences between American immigrant populations and their source countries far outweigh the similarities.
In order for MAGA to win, they need to convince America otherwise — they need to persuade you, the American citizen, that the fiction that undergirds their ideology is actually true. To this end, they need to get you to judge people in terms of their group, rather than as individuals. So they keep looking around for a group they think they can convince you to fear, to disdain, and ultimately to hate.
Remember last year, during the campaign season, when Trump and JD Vance declared that Haitian immigrants were eating people’s pets in Springfield, Ohio?
It was all B.S., of course. News crews descended on Springfield, but not even the most right-wing reporters could find a credible report of a single pet being eaten. JD Vance awkwardly begged the internet to “keep the cat memes flowing”, and never apologized for smearing a whole group of people, but at some point everyone realized it was a hoax.
That’s why you didn’t hear anything about cat-eating Haitian-Ohioans before the campaign season of 2024. And that’s why you haven’t heard anything about it since then. It wasn’t real; you were being played.
Now they’re trying again, with the Somalis of Minnesota. This time, they probably have a better shot at success. For one thing, Somalis in America are much poorer than their Haitian-American counterparts — Haitians in the U.S. have slightly below average income and average education levels, they commit few crimes, and they’re not prominent in politics. They’re basically just quiet middle-class people living pretty normal American lives.
Somalis, on the other hand, are an extremely poor group, with very high poverty rates and much lower income than Haitians, or immigrants in general; this is due to the fact that most of them are refugees or descendants of refugees, which are the least selected type of immigrants. Somalis are Muslim, unlike Haitians, which makes them both visually distinct (because of the hijab) and mentally associated with civilizational conflict. They’re not known for violence, but now they’re associated with Minnesota’s massive organized welfare fraud.
And unlike the Haitians of Ohio, the Somalis of Minnesota are prominent and powerful in local politics. They managed to do a sort of takeover of the Minneapolis Democratic Party, nominating one of their own, Omar Fateh, as the Democratic candidate over incumbent mayor Jacob Frey. Frey managed to beat Fateh in the general election, but only by appealing to a rival Somali clan and making flamboyant appeals to the Somali community.
This is hardly unprecedented in American politics — Irish immigrants built political machines that dominated the politics of many American cities in the 19th century. Given many decades, it’s likely that Somalis will assimilate, the same way the Irish did, and turn the organizational skills that allowed them to swindle the state of Minnesota and take over the Minneapolis Democrats to some more constructive use, like building drone factories (or whatever humans are doing 80 years from now).
But “many decades” is a very long time for Americans to wait in order not to worry about culture clash. And Americans aren’t used to urban ethnic machine politics these days,1 and the notion of an iconic American city being at the mercy of clan rivalries from one of the world’s poorest and most violent nations will naturally lend force to Trump’s argument that Somalis are trying to make Minnesota into another Somalia.
If Trump and MAGA succeed in getting a critical mass of regular Americans to reject Somalis categorically, as a racial group, then they win a crucial victory — not over the Somalis, who pose them no actual threat, but in terms of changing the terms of the discourse around race and immigration in America.
Once MAGA can convince you that “Are the Somalis bad?” is a legitimate question to ask, they then pretty much automatically get to ask the same question about every other group in America. They get to ask “Are Afghans bad?”, and “Are Haitians bad?”. They’ll get to ask “Are Jews bad?”, “Are Indians bad?”, and “Are Chinese people bad?”. Eventually they might even get around to asking “Are Italians bad?”, and so on. They will push as far as they can.
Even if those questions get answered in the affirmative — even if Italians and Indians and Haitians can all successfully defend their right to be in America by appealing to the court of MAGA opinion — the mere fact that they had to defend themselves as racial groups, instead of as individuals, will redefine what America is all about. It will move America toward being an estate society — a society where groups are accorded rights and privileges instead of individuals.
In the 20th century, American liberals successfully overcame all of the people who wanted to make the country a racial estate society — Jim Crow was outlawed, immigration laws were made (more or less) race-neutral, and so on. Liberals accomplished this by appealing to Americans’ deep-seated value of individualism — of the idea that people shouldn’t be judged by the group they were born into. That idea, captured most eloquently in MLK’s famous speech but repeated ad infinitum by leaders, writers, and activists, ultimately carried the day and made America the liberal nation I grew up in.
What I fear is that by embracing identity politics in the 2010s, progressives have thrown away liberals’ ultimate weapon. Appeals to individualism carry much less moral force when the people making those appeals just spent the last decade decrying colorblindness as a tool of systemic racism (or embracing people who made that claim).
This is not to say that rightists’ push to turn America into a balkanized racial hierarchy is progressives’ fault — it isn’t. Rightists are always trying to do this sort of thing; it’s not a reaction to anything progressives did. But there’s a reason this sort of racial collectivism was defeated and suppressed for a hundred years, and there’s a reason it’s breaking through now when it couldn’t before.
To be honest, they weren’t very relaxed about it in the 19th century either; anti-Irish sentiment resulted in vicious pogroms, gang wars, and whole newspapers devoted to spreading vicious anti-Irish rumors.
How did the screen you’re looking at right now get invented? There was a whole pipeline of innovation that started in the early 20th century. First, about a hundred years ago, a few weird European geniuses invented quantum mechanics, which lets us understand semiconductors. Then in the mid 20th century some Americans at Bell Labs invented the semiconductor. Some Japanese and American scientists at various corporate labs learned how to turn those into LEDs, LCDs, and thin-film transistors, which we use to make screens. Meanwhile, American chemists at Corning invented Gorilla Glass, a strong and flexible form of glass. Software engineers, mostly in America, created software that allowed screens to respond to touch in a predictable way. A host of other engineers and scientists — mostly in Japan, Taiwan, Korea, and the U.S. — did a bunch of incremental hardware improvements to make those screens brighter, higher-resolution, stronger, more responsive to touch, and so on. And voila — we get the screen you’re reading this post on.
This story is very simplified and condensed, but it illustrates how innovation is a pipeline. We have names for pieces of this pipeline — “basic research”, “applied research”, “invention”, “innovation”, “commercialization”, and so on — but these are approximate, and it’s often hard to tell where one of these ends and another begins. What we do know about this pipeline is:
It tends to go from general ideas (quantum mechanics) to specific products (a modern phone or laptop screen).
The initial ideas rarely if ever can be sold for money, but at some point in the chain you start being able to sell things.
That switch from non-monetizable to monetizable typically means that the early parts of the chain are handled by inventors, universities, government labs, and occasionally a very big corporate lab, while the later parts of the chain are handled mostly by corporate labs and other corporate engineers.
Very rarely does a whole chain of innovation happen within a single country; usually there are multiple handoffs from country to country as the innovation goes from initial ideas to final products.
Here’s what I think is a pretty good diagram from Barry Naughton, which separates the pipeline into three parts:
Over the years, the pipeline has changed a lot. In the old days, a lot of the middle stages — the part where theory gets turned into some basic prototype invention — were done by lone inventors like Thomas Edison or Nikola Tesla. Later, corporate labs took over this function, bringing together a bunch of different scientists and lots of research funding. Recently, corporate labs do less basic research (though they’re still very important in some areas like AI and pharma), and venture-funded startups have moved in to fill some of that gap.
The early parts of the pipeline changed too — university labs scaled up and became better funded, government labs got added, and a few very big corporate labs like Bell Labs even did some basic science of their own. The key innovation here was Big Science — in World War 2, America began using government to fund the early stages of the innovation pipeline with truly massive amounts of money. Everyone knows about the NIH and the NSF, but the really huge player here is the Department of Defense:
Japan, meanwhile, worked on improving the later parts of the chain. I recommend the book We Were Burning for a good intro to the ways that Japanese corporate labs utilized their companies’ engineering-intensive manufacturing divisions to make a continuous stream of small improvements to the final products, as well as finding ways to scale up and reduce costs (kaizen).
And finally, the links between the pieces of the pipeline — the way that technology gets handed off from one institution to another at different stages of the chain — changed as well. America passed the Bayh-Dole Act in 1980, making it a lot easier for university labs to commercialize their work — which thus made it easy and often lucrative for corporations to fund research at universities. (This had its roots in earlier practices by U.S. and German universities.)
Meanwhile, in parallel, the U.S. pioneered a couple of other models. There was the DARPA model, where an independent program manager funded by the government coordinates researchers from across government, companies, and universities in order to produce a specific technology that then gets handed off to both companies and the military. And there are occasional “Manhattan projects”, where the government coordinates a bunch of actors to create a specific technological breakthrough, like building nuclear weapons, landing on the moon, or sequencing the human genome.
So we’ve seen a number of big changes in the innovation pipeline over the years. And different countries have done innovation differently, adding crucial pieces and making key changes as their innovation ecosystems developed The UK pioneered the patent-protected “lone inventor” model (with some forerunners of modern venture capital). Germany created corporate labs and the research university. America invented Big Science, modern VC, and DARPA, while also scaling up modern university-private collaboration and undertaking a few Manhattan-type projects. And Japan added continuous improvement and continuous innovation at the end of the chain.
That story more or less brings us up from the 1700s to the late 2010s. That’s when China enters the innovation story in a big way.
China’s innovation boom
Up through the mid-2010s, China had a pretty typical innovation system — the government would fund basic research, companies would have labs that would create products, and so on. China wasn’t really at the technological frontier yet, though, so this system didn’t really matter that much for Chinese technology — most of the advances came from overseas, via licensing, joint ventures, reverse engineering, or espionage. If you’ve ever heard people talk about how China “steals” all its tech, they’re talking about this era — and “steal” means a whole bunch of different things.
In the 2010s, China’s growth slowed down. There were a lot of reasons for that, but one reason was that they were approaching the limits of how much technology they could transfer from overseas. They had to start inventing things on their own. So they did.
You’ve probably read a lot about Chinese innovation in the last few years. Most things you read will fall into one or more of three basic categories:
“Look how much money China is spending on research”
“Look how many academic papers China is publishing”
“Look which high-tech industries China is dominating”
And since salaries and materials and equipment are all cheaper in China, in PPP terms they’re actually spending a bit more on research than America now. And the gap is set to widen, with or without planned U.S. budget cuts:
As for scientific output, despite inflating their citation counts a lot with citation rings and other tricks, China now leads the world in high-quality STEM papers, especially in materials science, chemistry, engineering, and computer science:
And as for high-tech manufacturing, China is dominating there as well, except in a few narrow sectors where U.S. export controls have managed to keep key pieces of technology out of Chinese hands.
One other piece of evidence that China’s innovation is producing real results comes from the royalties that the world pays to Chinese companies to license their technologies. This amount has skyrocketed since China rolled out its new innovation system in the late 2010s, showing that China is producing lots of technology that the world is willing to pay for:
But although you’ll read a lot in the news about how much China is innovating, you almost never read a good explanation of how they’re doing it. Most people don’t seem to think about how research actually functions; people talk as if it’s just a black box where money goes in and cutting-edge high-tech products come out the other side. But it’s not a black box; the way that a country translates money into products is very important. It affects how productively the money will get used, who spends the money, how much can be deployed, what kinds of products and technologies that the system will create, and who will benefit from those products.
In fact, we know a lot about China’s innovation system — enough to know that in the last decade, they’ve created something new and powerful and interesting. If you want some readings, I strongly recommend:
Anyway, reading all this, it’s clear that like all the industrial nations before it, China has made big changes to the way innovation gets done. I’ll talk about what these changes are, and what they imply for the future of technology (and the economy), but first I think it’s useful to think a bit about the purpose of China’s innovation system.
_per_capita#/media/File:Map_of_countries_by_GDP_(PPP)_per_capita_in_2025.svg){kind=link}
_during_The_Deluge_and_Chmielnicki%27s_Uprising.png){kind=link}
