Oliver Kim reviews *How Africa Works*
That is the new book by Joe Studwell, my podcast with him should be coming out pretty soon. Here is Oliver’s new review. Excerpt:
Botswana is Studwell’s poster child for a successful democratic developmental coalition. (For this reason, it featured heavily in Acemoglu and Robinson’s Why Nations Fail as an example of “inclusive institutions”.)
Under the sound leadership of Seretse Khama, local chiefs were carefully co-opted at independence and the Botswana Democratic Party built up into a genuine national force. Khama also created a capable civil service, initially staffed by remaining Europeans, but gradually Africanized with sterling Batswana talent. This meant that when diamonds were discovered just around independence, the windfall was carefully managed, avoiding the worst effects of Dutch Disease. These mining revenues helped raise Botswana to upper middle-income status, making it the fourth-richest country in continental Africa.
Botswana’s chief failing, in Studwell’s view, was adhering too much to responsible policy orthodoxy—i.e., not enough industrial policy. There was no vision for large-scale industrialization, no coherent plan to create large numbers of factory jobs. Moreover, the political dominance of large cattle owners (Botswana was a society of pastoralists rather than farmers) meant that redistribution was never in the cards. The result is a relatively rich society, but one that is highly unequal.
You will be hearing my views on these issues soon enough. Oliver, of course, writes one of the very best Substacks in all of economics.
Optimal timing for superintelligence
There is a new paper by Nick Bostrom with that title:
Developing superintelligence is not like playing Russian roulette; it is more like undergoing risky surgery for a condition that will otherwise prove fatal. We examine optimal timing from a person-affecting stance (and set aside simulation hypotheses and other arcane considerations). Models incorporating safety progress, temporal discounting, quality-of-life differentials, and concave QALY utilities suggest that even high catastrophe probabilities are often worth accepting. Prioritarian weighting further shortens timelines. For many parameter settings, the optimal strategy would involve moving quickly to AGI capability, then pausing briefly before full deployment: swift to harbor, slow to berth. But poorly implemented pauses could do more harm than good.
Via Nabeel.
Thursday assorted links
1. Using Claude Code for academic work.
2. Younger Firms and CEOs Allow More Work from Home.
3. Extractive taxes were indeed a major force behind the French Revolution.
4. How much will “the human touch” persist?
5. “It was one attempt to do so, by Charles Jones of Stanford University, that entertained the negative top rate of -26%. If high earners produce a lot of ideas that help society, then “subsidising the discovery of new ideas through low tax rates may be as effective as redistribution in raising worker welfare”, he writes.” (The Economist)
6. Moral intuitions about love, romance, and reproduction are not Coasean.
7. Do not exercise options unless you have to!
I Regret to Inform You that the FDA is FDAing Again
I had high hopes and low expectations that the FDA under the new administration would be less paternalistic and more open to medical freedom. Instead, what we are getting is paternalism with different preferences. In particular, the FDA now appears to have a bizarre anti-vaccine fixation, particularly of the mRNA variety (disappointing but not surprising given the leadership of RFK Jr.).
The latest is that the FDA has issued a Refusal-to-File (RTF) letter to Moderna for their mRNA influenza vaccine, mRNA-1010. An RTF means the FDA has determined that the application is so deficient it doesn’t even warrant a review. RTF letters are not unheard of, but they’re rare—especially given that Moderna spent hundreds of millions of dollars running Phase 3 trials enrolling over 43,000 participants based on FDA guidance, and is now being told the (apparently) agreed-upon design was inadequate.
Moderna compared the efficacy of their vaccine to a standard flu vaccine widely used in the United States. The FDA’s stated rationale is that the control arm did not reflect the “best-available standard of care.” In plain English, that appears to mean the comparator should have been one of the ACIP-preferred “enhanced” flu vaccines for adults 65+ (e.g., high-dose/adjuvanted) rather than a standard-dose product.
Out of context, that’s not crazy but it’s also not necessarily wise. There is nothing wrong with having multiple drugs and vaccines, some of which are less effective on average than others. We want a medical armamentarium: different platforms, different supply chains, different side-effect profiles, and more options when one product isn’t available or isn’t a good fit. The mRNA vaccines, for example, can be updated faster than standard vaccines, so having an mRNA option available may produce superior real-world effectiveness even if it were less efficacious in a head-to-head trial.
In context, this looks like the regulatory rules of the game are being changed retroactively—a textbook example of regulatory uncertainty destroying option value. STAT News reports that Vinay Prasad personally handled the letter and overrode staff who were prepared to proceed with review. Moderna took the unusual step of publicly releasing Prasad’s letter—companies almost never do this, suggesting they’ve calculated the reputational risk of publicly fighting the FDA is lower than the cost of acquiescing.
Moreover, the comparator issue was discussed—and seemingly settled—beforehand. Moderna says the FDA agreed with the trial design in April 2024, and as recently as August 2025 suggested it would file the application and address comparator issues during the review process.
Finally, Moderna also provided immunogenicity and safety data from a separate Phase 3 study in adults 65+ comparing mRNA-1010 against a licensed high-dose flu vaccine, just as FDA had requested—yet the application was still refused.
What is most disturbing is not the specifics of this case but the arbitrariness and capriciousness of the process. The EU, Canada, and Australia have all accepted Moderna’s application for review. We may soon see an mRNA flu vaccine available across the developed world but not in the United States—not because it failed on safety or efficacy, but because FDA political leadership decided, after the fact, that the comparator choice they inherited was now unacceptable.
The irony is staggering. Moderna is an American company. Its mRNA platform was developed at record speed with billions in U.S. taxpayer support through Operation Warp Speed — the signature public health achievement of the first Trump administration. The same government that funded the creation of this technology is now dismantling it. In August, HHS canceled $500 million in BARDA contracts for mRNA vaccine development and terminated a separate $590 million contract with Moderna for an avian flu vaccine. Several states have introduced legislation to ban mRNA vaccines. Insanity.
The consequences are already visible. In January, Moderna’s CEO announced the company will no longer invest in new Phase 3 vaccine trials for infectious diseases: “You cannot make a return on investment if you don’t have access to the U.S. market.” Vaccines for Epstein-Barr virus, herpes, and shingles have been shelved. That’s what regulatory roulette buys you: a shrinking pipeline of medical innovation.
An administration that promised medical freedom is delivering medical nationalism: fewer options, less innovation, and a clear signal to every company considering pharmaceutical investment that the rules can change after the game is played. And this isn’t a one-product story. mRNA is a general-purpose platform with spillovers across infectious disease and vaccines for cancer; if the U.S. turns mRNA into a political third rail, the investment, talent, and manufacturing will migrate elsewhere. America built this capability, and we’re now choosing to export it—along with the health benefits.
That Was Then/This is Now
Hat tip: Logan Dobson.
The economics of mass deportation
Following the removal of 50% of unauthorized immigrants, in the short run average native real wages rise 0.15% nationally, driven by an increase in the capital-labor ratio. In the long run, however, native real wages fall in every state, and by 0.33% nationally, as capital gets decumulated in response to a lower population. Consumer prices in the sectors intensive in unauthorized workers – such as Farming – rise by about 1% relative to the price of the average consumption basket, while most other sectors experience negligible relative price changes.
That research result is from
Past Automation and Future A.I.: How Weak Links Tame the Growth Explosion
From Charles I. Jones and Christopher Tonetti:
How muchof past economic growth is due to automation, and what does this imply about the effects of A.I. and automation in the coming decades? We perform growth accounting using a task-based model for key sectors in the U.S. economy. Historically, TFP growth is largely due to improvements in capital productivity. The annual growth rate of capital productivity is at least 5pp larger than the sum of labor and factor-neutral productivity growth. The main benefit of automation is that we use rapidly-improving machines instead of slowly-improving humans on anincreasing set of tasks. Looking to the future, we develop an endogenous growth model in which the production of both goods and ideas is endogenously automated. We calibrate this model based on our historical evidence. Two key findings emerge. First, automation leads economic growth to accelerate over the next 75 years. Second, the acceleration is remarkably slow. By 2040, output is only 4% higher than it would have been without the growth acceleration, and by 2060 the gain is still only 19%. A key reason for the slow acceleration is the prominence of “weak links” (an elasticity of substitution among tasks less than one). Even when most tasks are automated by rapidly improving capital, output is constrained by the tasks performed by slowly-improving labor.
And an important sentence from the paper itself:
…, the key gain from automation is that it allows production of a task to shift away from slowly-improving human labor to rapidly-improving machines.
The authors stress that those are preliminary results, and the numbers are likely to change. For the pointer I thank the excellent Kurtis Hingl, who is also my research assistant.
*The Almighty Dollar*
The author is Brendan Greeley, and the subtitle is 500 Years of the World’s Most Powerful Money. A very well-timed book, excellent on the history of the dollar as it spans the centuries, I was happy to write a blurb for it.
Wednesday assorted links
1. The rise of the Saudi-UAE split.
2. How Peru came the leading exporter of blueberries.
3. One view on how to respond to AI potentially taking your job.
4. Austin Steady rates books mentioned on MR.
5. 36 hours in Lagos (NYT).
6. Why economists linearize everything.
My New Jersey history podcast with “Exit Interviews”
Exit Interviews is a new podcast run by David Piegaro. I am honored to be one of the first few guests, along with Chris Christie. Think of this session as “Tyler Cowen as regional thinker.” Almost 100% fresh material, not to mention some trolling directed at Central and South Jersey, Philly too. Here is my episode.
Definitely recommended, and let us hope that David Remnick gets on soon to defend the honor of River Vale vs. Hillsdale in Bergen County…
Immigration and health for elderly Americans
We measure the impact of increased immigration on mortality among elderly Americans, who rely on the immigrant-intensive health and long-term care sectors. Using a shift-share approach we find a strong impact of immigration on the size of the immigrant care workforce: admitting 1,000 new immigrants would lead to 142 new foreign healthcare workers, without evidence of crowd out of native health care workers. We also find striking effects on mortality: a 25% increase in the steady state flow of immigrants to the US would result in 5,000 fewer deaths nationwide. We identify reduced use of nursing homes as a key mechanism driving this result.
That is from a new NBER working paper by
My simple model of fertility decline
My core model is both simple and depressing. Fertility rates have declined around the world because birth control technologies became much better and easier to use. And people — women in particular — just do not want that many kids.
I do understand that better birth control happened a long time ago, for instance birth control pills become widely available in the wealthier countries in the 1960s, or sometimes the 1970s. Nonetheless the diffusion of new technologies can be very slow, and for norms to shift it can take generational turnover or even a bit more. Plus “fertility contagion effects” take a long time to work their way fully through the system.
Those long lags may be difficult to swallow, but social science has numerous examples of very long operative mechanisms. (Just think of how long it took potential migrants to exploit open borders, for instance pre-WWI.) Furthermore, fertility rates have indeed been falling for a long time in the wealthier countries.
So a lot of women, once they face the realities of the stress and trying to make ends meet, want only one kid. You end up with a large number of one kid families, some people who never marry/procreate at all, and a modest percentage of families with 2-4 kids. There are also plenty of cases cases where the guy leaves, self-destructs, or never marries, after siring a single child with a woman. That gives you the fertility rates we are seeing, albeit with cultural and economic variation.
Richard Hanania considers why income is not the driving force behind the decline, and why the decline is continuing.
Part of this model is that many women just love having a child. They love “children” so much that a single child fills up their needs and desires.
I see a similar mechanism in my own life. I very much enjoy having Spinoza around the house, but I have zero desire to take in another canine. Whenever I want more “dog attention,” I can assure you that the supply is highly elastic. Similarly, a single kid can take up a lot of your time and affection, again supply is elastic from the side of the kid. Maybe parents learning how much they can enjoy a single kid has been another cultural lag?
Under my preference-driven model, fertility declines are very difficult to reverse. I believe that is also consistent with the evidence to date.
So this is a problem we need to worry about. The asymptote is rather unpleasant, and the path along the way involve less human well-being, possibly less innovation, and maybe some major fiscal crises as well.
As Arnold Kling would say, “Have a nice day.”
Recursive self-improvement from AI models
With Claude Opus 4.6 and 5.3 Codex, both stellar achievements, the pace is heating up:
OpenAI went from its last Codex release, on December 18, 2025, to what is widely acknowledged to be a much more powerful one in less than two months. This compares to frequent gaps of six months or even a year between releases. If OpenAI can continue at that rate, that means we can easily get four major updates in a year.
But the results from what people in the AI world call “recursive self-improvement” could be more radical than that. After the next one or two iterations are in place, the model will probably be able to update itself more rapidly yet. Let us say that by the third update within a year, an additional update can occur within a mere month. For the latter part of that year, all of a sudden we could get six updates—one a month: a faster pace yet.
It will depend on the exact numbers you postulate, but it is easy to see that pretty quickly, the pace of improvement might be as much as five to ten times higher with AI doing most of the programming. That is the scenario we are headed for, and it was revealed through last week’s releases.
Various complications bind the pace of improvement. For the foreseeable future, the AIs require human guidance and assistance in improving themselves. That places an upper bound on how fast the improvements can come. A company’s legal department may need to approve any new model release, and a marketing plan has to be drawn up. The final decisions lie in the hands of humans. Data pipelines, product integration, and safety testing present additional delays, and the expenses of energy and compute become increasingly important problems.
And:
Where the advance really matters is for advanced programming tasks. If you wish to build your own app, that is now possible in short order. If a gaming company wants to design and then test a new game concept, that process will go much faster than before. A lot of the work done by major software companies now can be done by much smaller teams, and at lower cost. Improvements in areas such as chip design and drone software will come much more quickly. And those advances filter into areas like making movies, in which the already-rapid advance of AI will be further accelerated.
Here is more from me at The Free Press.
Tuesday assorted links
1. Becca Rothfeld and Jordan Salama joining The New Yorker.
2. Discord will be requiring a face scan or ID for full access. I wonder what induced that change?
3. Meta-study showing that the returns to education are not so high.
4. Is she the most important philosopher in the world? (WSJ)
5. Do not exercise options unless you have to!
The politics of using AI
Using new data from the Gallup Workforce Panel, we document a persistent partisan gap in self-reported AI use at work: Democrats are consistently more likely than Republicans to report frequent use. In 2025:Q4, for example, 27.8% of Democrats report using AI weekly or daily, compared with 22.5% of Republicans. Democrats also report deeper task-level integration, using AI in 16% more work activities than Republicans. Consistent with this, Democrats are employed in occupations with higher predicted AI exposure based on task-content measures and report larger perceived differences in AI-related job displacement risk. However, in regression models the partisan gap in AI use disappears once we control for education, industry, and occupation, indicating that observed differences primarily reflect compositional variation rather than political affiliation per se.
That is from a new paper by Nicholas Bloom and Christos Makridis.