Category: Economics
India AI Data MCP
The Government of India’s Ministry of Statistics and Program Implementation has created an impressive Model Context Protocol (MCP) to connect AI’s to Indian datasets. An AI connected to data via an MCP essentially knows the entire codebook and can make use of the data like an expert. Once connected one can query the data in natural language and quickly create graphs and statistical analysis. I connected Claude to the MCP and created an elegant dashboard with data from India’s Annual Survey of Industries. Check it out.
My excellent Conversation with Joe Studwell
Here is the audio, video, and transcript. The conversation is based around Joe’s new and very good book How Africa Works: Success and Failure in the World’s Last Developmental Frontier. Here is part of the episode summary:
Tyler and Joe explore whether population density actually solves development, which African countries are likely to achieve stable growth, whether Africa has a manufacturing future, why state infrastructure projects decay while farmer-led irrigation thrives, what progress looks like in education and public health, whether charter cities or special economic zones can work, and how permanent Africa’s colonial borders really are. After testing Joe’s optimism about Africa, Tyler shifts back to Asia: what Japan and South Korea will do about depopulation, why industrial policy worked in East Asia but failed in India and Brazil, what went wrong in Thailand, and what Joe will tackle next.
Excerpt:
COWEN: Does Africa have a manufacturing future? Is robotics coming, AI, possibly some reshoring?
STUDWELL: Yes. I believe that Africa does have a manufacturing future.
COWEN: But making what? And at what cost of energy?
STUDWELL: They will start, as everybody does, producing garments, producing textiles, which in certain enclaves is already going on in Madagascar, in Lasutu, in Morocco, and they’ll move on to other things. They’ll start with those things because they are the most labor cost-sensitive products.
Africa is now in a position where — depending on which state you’re looking at, and taking China as a reference point — the cost of labor is now between a half and one-tenth of what it is in China. Factory labor is now around $600 a month at its cheapest. In a country like Ethiopia or Madagascar, it’s $60 or $65 a month. So, it’s a 10th of the cost, and that’s already beginning to have a bit of effect, often with Chinese firms moving production to Africa.
So, I think there is a future for manufacturing. It will depend on the extent to which African governments understand that you don’t really move forward fast for very long without manufacturing, that every developed country — apart from a few petro states and financial centers — has gone through a manufacturing phase of development. It depends on the extent to which African governments engage with that, but some, without doubt, will.
The Ethiopians, for instance, have already attempted to do that. What they’re trying to do has been somewhat derailed by the two-year civil war that took place from 2020, but they’re back on it now, and they’re trying to move forward.
The idea that robotics and AI are going to change the story I personally do not buy, principally for two reasons. One is the cost reason, because whenever people talk about what’s happening with robotics, no one ever talks about the cost of robots. In garmenting, for instance, even a basic robot will cost you in excess of $100,000, and you pay the cost upfront, and you’ve then paid that, whether there’s demand for your products or not. Also, in garmenting and in textiles, robots don’t work very well because they can’t work with material very well. They’re much better at working with solid things.
So, you’ve spent $100,000 for a robot when you can go out in somewhere like Tana in Madagascar and get another skilled — because they’ve been doing it now for 20 years — garmenting employee for $60 or $65 to make the new order that you just got. And if the order doesn’t come through, you can sack them. You see what I’m saying? There’s a point about the cost of robotics.
COWEN: But think of automation more generally — it’s not that expensive. Most countries are de-industrializing. Even South Africa has been de-industrializing for a while, and China maybe has peaked out at industrialization, measured in terms of employment. It’s hard to trust their numbers. But maybe just everywhere is going to deindustrialize, and that will be very bad for Africa.
STUDWELL: I don’t think so. I think South Africa is deindustrializing because the ANC has followed a hyper-liberal approach to economic policy. I don’t think the ANC has ever really understood economic policy, frankly, so South Africa is an outlier in that respect. There are many other states in Africa, whether Nigeria or Ethiopia, which understand they’ve got to have a manufacturing future and intend to pursue one.
Then, as I was saying, the other point is, what people miss is the flexibility with robotics and AI. There’s very limited flexibility with robotic and automated production. When demand goes up, you can’t just stick in more robots, but when demand goes up in a people-operated factory, where the cost of labor is low, you can stick in more people and produce more.
Just one example: during COVID, when everybody was having home deliveries of supermarket goods, the price of a UK firm called Ocado, which runs a supermarket, but was also developing the software and consulting around building blind warehouses went up through the roof, but now it’s down through the floor.
And only last week, Kroger supermarket in the US said, “We’re closing five of these super-modern blind warehouses.” And the reason, fundamentally, is because they lack the flexibility that human labor brings to the job. So, I’m not saying that robots, automation, and AI are not important. They are important. What I am saying is that they are not going to derail a manufacturing future for a number of African countries that aggressively pursue it.
COWEN: But there’re a lot of developing nations around the world — you could look at India, you could look at Pakistan, even Thailand — where manufacturing has not taken off the way one might have wanted. There’re just major forces operating against it. And in the US, manufacturing employment was once 37 percent of the workforce; now it’s 7 percent to 8 percent.
It just seems like it’s swimming upstream for Africa — which again, has quite expensive energy — to think it will do that well. And again, South Africa had very good technology, pretty high state capacity. I don’t see the alternate world state where a wiser ANC would have made that work.
STUDWELL: Well, oddly enough, before the end of Apartheid, the manufacturing performance of South Africa was really not bad at all, with classic industrial policy, quite high levels of protection, and so forth. I think that demand for manufactured goods will continue to be high around the world, and the labor cost will continue to be a prime determinant of where producers go for low value-added goods. So, I think that the opportunity is there for African countries.
COWEN: But say there’re transportation costs internally, energy costs, political order uncertainty. Where’s the place where people really want to put all these manufacturing firms?
Interesting throughout, recommended.
The Cassidy Report on the FDA
Senator Bill Cassidy (R-La.) released a new report on how to modernize the FDA. It has some good material.
… FDA’s process for reviewing new products can be an unpredictable “black box.” FDA teams can differ greatly in the extent to which they require testing or impose standards that are not calibrated to the relevant risks. The perceived disconnect between the forward leaning rhetoric and thought leadership of senior FDA officials and cautious reviewer practice creates further unpredictability. This uncertainty dampens investment and increases the time it takes for patients to receive new therapies.
Companies report that they face a “reviewer lottery,” where critical questions hinge on the approach of a small number of individuals at FDA. Some FDA review teams are creative and forward-leaning, helping developers design programs and overcome obstacles to get needed products to patients, without cutting corners. FDA’s Oncology Center of Excellence (OCE), for example, is repeatedly identified as a model for providing predictable yet flexible options for bringing new drugs to cancer patients. OCE is now a dialogue-based regulatory paradigm that has facilitated efforts by academia, industry, the National Institutes of Health (NIH), and others to develop new cancer therapies and launch innovative programs and pilots like Project Orbis, RealTime Oncology Review.
That’s quite right. In my report with DiMasi and Milne, we show that review times vary markedly within the FDA, across divisions, without clear cause. Put aside the big questions of FDA or no FDA–scaling the practices of the better-performing divisions would likely yield large gains.
Mike Doherty further asks the AI what Alex Tabarrok would think of the Cassidy report. It replies:
Strongest features: It correctly diagnoses FDA as a bottleneck to innovation, emphasizes “least burdensome” evidence, calls for streamlined pathways (505(b)(2)-style for biologics, biosimilar simplification), and worries about U.S. losing early-phase trials to other countries. This is very much in Tabarrok’s wheelhouse: regulation as a hidden tax on innovation and growth.
Weakest features: It rarely quantifies tradeoffs, almost never confronts the “invisible graveyard” of patients who never see therapies because of delay, and treats reforms as marginal tweaks rather than testing fundamentally different regulatory models (e.g., insurance-based approval, private certification, or sunset/experimentation with parallel regimes).
If you imagine this as a draft memo handed to Tabarrok, he’d likely say: “Good directionally; now add 50% more economics, 50% more quantification, and 100% more willingness to experiment with institutional competition.”
Yeah, pretty good.
Addendum: In other FDA news see also Adam Kroetsch on Will Bayesian Statistics Transform Trials?
Addendum 2: FDA has now agreed to review Moderna’s flu vaccine which is good although the course reversal obviously speaks to the unpredictability of the FDA.
A simple test of how immigration really is going
I suggest looking at whether real estate prices in a particular locale have been rising or falling. If immigration is “ruining” a particular city, we would expect homes and other property values in that place to become much cheaper.
Home values have historically served as a strong indicator of the health of a city. Consider Detroit. It was one of the premier American cities in the mid-20th century, but the region lost a lot of its automobile industry to foreign competition, and crime rose precipitously. The city also was poorly managed. The result in real estate markets was a collapse in prices. If anyone asked you to point to quantifiable evidence for the decline in Detroit, it was easy to do so.
Detroit has undergone a renaissance since its nadir. New businesses have opened, crime rates have fallen, and the city feels more lively again. And since that turn of fortune, often dated around the 1990s, Detroit real estate has made a major comeback, putting aside the price collapse of the Great Recession in 2008. Home prices are not a perfect measure of how the city is doing, but they do pick up major and radical trends, both on the downside and on the upside.
The nice thing about market prices is that they show how buyers weigh the benefits of immigration against costs. Say some new immigrants have moved into your community and the quality of the schools has declined somewhat and traffic is modestly worse. At the same time, there are new businesses, the streets feel more lively, and it is easier to get a good local plumber. In the abstract, it is hard to tell which effects might be most important. But individuals, when bidding for homes or deciding to sell, make their own judgments. What happens to the home prices is a reflection of the collective judgments of people with major decisions about their lives on the line.
Of course in most of the Western world, including Malmo, real estate prices are healthy and very often rising. Here is the full Free Press link, by yours truly. The piece of course does cover the usual caveats, such as bubbles and busts, but note NIMBY factors will not alone reverse the basic conclusions.
“You see tech and AI everywhere but in the productivity statistics”
How many times have I heard versions of that claim? Erik Brynjolfsson picks up the telephone in the FT:
While initial reports suggested a year of steady labour expansion in the US, the new figures reveal that total payroll growth was revised downward by approximately 403,000 jobs. Crucially, this downward revision occurred while real GDP remained robust, including a 3.7 per cent growth rate in the fourth quarter. This decoupling — maintaining high output with significantly lower labour input — is the hallmark of productivity growth.
My own updated analysis suggests a US productivity increase of roughly 2.7 per cent for 2025. This is a near doubling from the sluggish 1.4 per cent annual average that characterised the past decade.
It is fine to suggest caution in interpreting such statistics, but they hardly push the other way.
Minimum Wages for Gig Workers Can’t Work
In 2017, I analyzed the Uber Tipping Equilibrium:
What is the effect of tipping on the take-home pay of Uber drivers? Economic theory offers a clear answer. Tipping has no effect on take home pay. The supply of Uber driver-hours is very elastic. Drivers can easily work more hours when the payment per ride increases and since every person with a decent car is a potential Uber driver it’s also easy for the number of drivers to expand when payments increase. As a good approximation, we can think of the supply of driver-hours as being perfectly elastic at a fixed market wage. What this means is that take home pay must stay constant even when tipping increases.
…If Uber holds fares constant, the higher net wage (tips plus fares) will attract more drivers but as the number of drivers increases their probability of finding a rider will fall. The drivers will earn more when driving but spend less time driving and more time idling. In other words, tipping will increase the “driving wage,” but reduce paid driving-time until the net hourly wage is pushed back down to the market wage.
A paper by Hall, Horton and Knoepfle showed that’s exactly what happened.
More recently, in 2024, Seattle implemented “PayUp”, a pay package for gig workers like DoorDash workers that required a minimum wage based on the time worked and miles travelled for each offer. Note that this is not a minimum wage for all workers but for one type of worker in a large market. For this reason, we can use the same analysis as with Uber tipping. The supply of workers is very elastic and essentially fixed at the market wage for workers of similar skill. Thus, we would expect a zero effect on net pay.
Here is a recent NBER paper by An, Garin and Kovak looking at the effects of the Seattle law:
We find that the minimum pay law raised delivery pay per task….At the same time, the policy led to a reduction in the number of tasks completed by highly attached incumbent drivers (but not an increase in exit from delivery work), completely offsetting increased pay per task and leading to zero effect on monthly earnings. We find evidence that drivers experienced more unpaid idle time and longer distances driven between tasks…Using a simple model of the labor market for platform delivery drivers, we show that our evidence is consistent with free entry of drivers into the delivery market driving down the task-finding rate until expected earnings return to their pre-reform level.
All of this is a general result of the Happy Meal Fallacy.
Malthus had real influence
From a recent paper by Eric Robertson:
Public officials often fail to implement government policy as directed, yet the role of economic ideas in shaping these implementation choices is poorly understood. This paper provides causal evidence that exposure to economic ideas can durably influence bureaucrat behavior. I study British colonial bureaucrats in India, exploiting a natural experiment created by the abrupt death of Thomas Malthus in 1834, replacing his economics instruction at a bureaucrat training college for that of a contemporary critic, Richard Jones. Whereas Malthus regarded economic distress as a natural mechanism for restoring equilibrium by reducing population growth, Jones disagreed with this view. Linking rainfall shocks to district-level fiscal responses, I show that officials trained by Malthus delivered less relief during droughts, providing 0.10-0.25 SD less aid across all major measures compared with officials taught by Jones. The results reveal that exposure to abstract economic ideas can shape real-world policy implementation for decades.
This may be a case where using rainfall shocks in a paper actually makes sense. Via Krzysztof Tyszka-Drozdowski.
Natural and Artificial Ice
Excellent Veritasium video on the 19th century ice industry. Shipping ice from America to India would hardly seem like a wise idea—it’s hard to imagine ever getting a committee to approve such a venture—but entrepreneurs are free to try wacky ideas all the time, and sometimes they pay off, resulting in great riches. That’s the story of the “Ice King,” Frederic Tudor, who lost money for years before figuring out the insulation and logistics needed to make the trade profitable.
What I hadn’t fully appreciated is how the ice trade reshaped shipping, diet, and city design before the invention of mechanical refrigeration. Ice created the cold chain, and the cold chain made it possible to move fresh meat, fish, and produce over long distances. That in turn enabled cities to grow far beyond what local agriculture could support and shifted the American diet from salted and smoked provisions toward fresh food.
The profits of the ice trade encouraged investment in artificial ice which initially was met with resistance—natural ice is created by God!—a classic example of incumbents wrapping their economic interests in moral language, a pattern we see repeated with every disruptive technology from margarine to ridesharing.
Lots of lessons in the video about option value, permissionless innovation, and creative destruction. New technologies destroy old industries and create new ones that no one could have foreseen. The moral panic over artificial ice replacing the natural kind is no doubt familiar.
Hat tip: Naveen Nvn
The economics of corporate espionage
Weprovide systematic evidence on the economic damages from espionage to US firms and industries. Compiling a comprehensive dataset of publicly disclosed espionage incidents from 1995-2024, we establish that espionage has substantial negative effects on targeted f irms. In an event-study design, revenues and R&D expenditures at targeted firms decline by roughly 40% within five years, with effects persisting for up to a decade. These effects do not appear for firms unsuccessfully targeted for espionage, supporting a causal interpretation. These firm-level damages translate into measurable aggregate effects on US industry: exports in targeted sectors decline by 60% over a decade. Given these substantial damages, we investigate whether firms restrict knowledge sharing in response to espionage. Across a wide range of outcomes, we find no evidence of such restrictions. Firms do not reduce their patenting with foreign inventors, and do not discriminate in employment based on perceived espionage risk. Overall, espionage has clear economic harms to targeted firms and US industry, but firms are puzzlingly unresponsive in how they manage innovation.
That is from a new paper by Andrew Kao and Karthik Tadepalli. Via Kris Gulati.
Taxing Beta, Exempting Alpha: A Benchmark-Based Inheritance Regime
This paper proposes a generational benchmark inheritance regime as a structural replacement for the federal estate tax. By distinguishing between systemic market returns (Beta) and active value creation (Alpha), the regime captures the passive growth of capital at generational boundaries while fully exempting idiosyncratic surplus. Using a Pareto tail interpolation (α ≈ 1.163) calibrated to Federal Reserve wealth data, we estimate baseline annual revenue of approximately $295 billion under conservative assumptions. This revenue is sufficient to finance a 2.1 percentage point reduction in the OASDI payroll tax, shifting the fiscal burden from labor to underperforming dynastic capital. Unlike continuous wealth taxes, the regime requires no new valuation machinery, relying exclusively on existing estate and gift tax procedures. We situate the proposal within the Jeffersonian principle of usufruct and the modern literature on optimal inheritance taxation.
Minimum wage hikes and robots
This paper studies how minimum wage policy affects firms’ adoption of automation technologies. Using both state-level measures of robot exposure and novel plant-level data on industrial robot imports linked to U.S. Census microdata from 1992-2021, we show that increases in minimum wages raise the likelihood of robot adoption in manufacturing. Our preferred identification exploits discontinuities at state borders, comparing otherwise similar firms exposed to different wage floors. Across specifications, a 10 percent increase in the minimum wage increases robot adoption by roughly 8 percent relative to the mean.
That is from Erik Brynjolfsson, et.al., including Andrew Wang. Via the excellent Kevin Lewis.
By the way, a photo from our textbook Modern Principles of Economics:
Changes in the Gender Wage Gap for Business Professionals
In the United States, much of the gap in earnings between men and women is due to the persistent gap for high wage earners. This paper explores changes in the gender wage gap for MBAs graduating from a large public university over 30 years. We document large gender wage gaps on average, which grow in the course of men’s and women’s careers. Comparing graduates at identical career stages across time periods to address composition concerns, we show that the raw gender wage gap has shrunk by 33 to 50 percent over the last two decades. Additionally, the temporal pattern of the gap has fundamentally shifted: while gaps only emerged over time in earlier decades, significant gaps now emerge immediately. Convergence in labor supply factors, particularly hours worked, explains much of the narrowing gap, alongside shifts in industry composition. However, unexplained wage gaps persist for recent graduates from the very start of their careers, suggesting different underlying mechanisms across cohorts. These findings highlight both progress in gender wage equity among business professionals and concerning patterns that emerge earlier in careers than in previous decades.
That is from a recent NBER working paper by
The import of cross-task productivity
Given that LLMs seem to be able to automate so many small tasks, why don’t we see large productivity effects?
I drafted a short paper recently exploring the possibility that it’s for the same reason (or at least one of the reasons) that labor is typically bundled into multi-task jobs, instead of transacted by the task, in the first place: because performing a task increases one’s productivity not only at the task itself but at related tasks.
For example, say you used to spend half your time coding and half your time debugging, and the LLM can automate the coding but you still have to do the debugging. If you’re more productive at debugging code you write yourself, this (1) explains why “coder” and “debugger” aren’t separate jobs, and (2) predicts that the LLM won’t save half your time. If you’re half as productive at debugging code you didn’t write, or less, the LLM saves you no time at all.
So I was excited to see @judyhshen and @alextamkin’s paper from a week or two ago finding basically just that!
At least the way I’m thinking about it, “cross-task learning” should make the productivity impacts of automating tasks more convex: – Automating the second half of a job should be expected to have much more of an impact than automating the first half; and – If the machines can learn from their and each others’ experience, as a worker learns by doing from her own experience, then automating two jobs will have more than twice the impact of automating one.
That is from Philip Trammell. Here is his short piece. Here is the Shen and Tamkin paper. This is all very important work for why the AI growth take-off will be much slower than the power of the models themselves might otherwise indicate. The phrase “…and then all at once” nonetheless applies. But when?
These short pieces and observations are likely among the most important outputs economists will produce this year. But are they being suitably rewarded?
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.
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.