Extraordinary Machine

I am joining the war on Daron Acemoglu on the side of Acemoglu

daron acemoglu happy moments compilation (00:10)

I can't help it, the road just rolls out behind me
Be kind to me, or treat me mean
I'll make the most of it, I'm an extraordinary machine

Fiona Apple, “Extraordinary Machine”

Recently, The Economist has put out an article titled “The world’s most influential economist is oddly unconvincing” about the 2024 Nobel Prize in Economics winner Daron Acemoglu. He’s a frequent source of discussion on this blog, so I thought that I’d give my two cents on the topic. What is the core of The Economist’s attack on Daron Acemoglu? There’s, roughly speaking, two avenues: his work on insitutions, and his work on AI. The Economist piece on Acemoglu takes two major avenues of attack: that his work on institutions is empirically shoddy, and that it’s theoretically imprecise.

Mostly Harmful Econometrics

The methodological critique is the same one I’ve made a bunch of times: that the data on the 2001 Acemoglu, Johnson, and Robinson paper that made his name as an economist has low quality and is very dependent on hard to defend decisions around how to impute data.

In particular, the “The Colonial Origins of Comparative Development” (2001) has a simple strategy to prove institutions, not geography, cause development: because better institutions and better GDP are also correlated to variables that aren’t measurable, you need a middleman between the two. This “middleman”, called an instrument, is correlated only to the cause variable (institutions) and not to either the consequence directly (GDP) or to other unobservables that might be mucking up the correlation (culture, for instance). The instrument is then used as a proxy for the cause, so it has to be very tightly correlated to the variable it’s a proxy for, and that it cannot be correlated to any other variable under consideration.

The instrument they use is the mortality of European settlers: different places had different geographic conditions, which are half-convincingly shown to be uncorrelated to the other variables, and those conditions made it so that settling in some places was pleasant, and unpleasant in others. Unpleasant locations needed more slaves, which led to worse institutions; meanwhile, nicer locations attracted more European workers, who were able to eventually defend their legal and property rights. The empirical work shows, unsurprisingly because the paper got published, that this relationship holds. The Economist says:

His Nobel rests heavily on a paper released in 2001 with Messrs Johnson and Robinson, which researchers have cited more than 23,000 times. It seeks to explain why some countries are rich and others poor (…) The paper encouraged wonks to think more deeply about institutions, but they have also picked it apart. The Nobel committee’s report in 2024 conceded that the mortality data are “sometimes sketchy”.

Why is the data sketchy? Well, it’s one of the most famous topics of academic debate in recent years. In 2012, David Albouy verified whether the 2001 paper was too dependent on imputed mortality rates - that is, if Guatemala doesn’t have mortality, then you can use Nicaragua’s. Harmless enough, but the paper also extrapolates the mortality of say, bishops, or traders, or soldiers, to the mortality of all settlers: which, you could argue, is a relationship itself, but this leads to conflicting evidence in some cases. Albouy, thus, notes that if you control for these issues, the relationship stops holding; however, The Economist quotes Acemoglu: Albouy had to drop half his sample to make the controls work, which means that the most developed countries (the USA, Australia, and Canada) weren’t included. A recent blog post by Joe Francis looks at the claims by both Albouy and Acemoglu & co. (they, unsurprisingly, wrote a rebuttal to the rebuttal that, also unsurprisingly, wasn’t very convincing), and finds that Albouy just makes the stronger case: the results collapse when you don’t include the three or four countries (the ones mentioned above) that were colonized by Europeans for an extensive period of time and actually became rich, and hold when you drop Africa, the continent with the worst data and also, not very coincidentally, the worst outcomes.

Econometrically speaking, there appears to be a massive problem with the instrument: the estimate for the institutions to GDP correlation is smaller than the one for the mortality to GDP one; this means that, rather than taking out the problematic unobservables, the instrument is adding more of them, because mortality is linked to geography and geography is linked to other variables: for example, if oil-rich places were also likelier to have higher rates of malaria, you’d have a bunch of very rich, very extractive countries with very high settler mortality. And similarly, the history behind the colonial mortality is a bit of a problem: different places were colonized at different times; for instance, nearly as much time passed between the conquest of Mexico and the Scramble for Africa as between the Age of “Discovery” and the Norman Conquest. When Europe conquered Latin America, capitalism simply did not exist yet, and developed during the time period where the Americas were colonies. During this time, also, Latin American economic institutions impacted African institutions because that’s where Europe took the slaves from: importantly, one of the crucial conditions for statistical inference is the Stable Unit Treatment Value Assumption or SUTVA. According to SUTVA, the two core condiitons to assess impact are that all units receive the same “treatment” (e.g. if you’re evaluating vaccinations then it has to always be the same vaccine) and, most importantly, there are no externalities between treated and untreated units. This means that, before country X was colonized, it cannot have the impacts of colonial institutions; instead, what we see based on the slave trade is that there are spillovers. To quote Francis on the topic of SUTVA in AJR (2001), “The low levels of settler mortality meant that the European offshoots received the “good institutions” of the metropole. This was their treatment, while the other colonies act as the counterfactual. The problem is that the other colonies were themselves affected by those “good institutions.” Both the European offshoots and the other colonies were embedded in the same network of world trade.”

But this shows the major problem here: the quality, and quantitative facts, of the data is correlated with institutional quality. For example, in The Better Angels of Our Nature, Stephen Pinker at some point cites the An Lushan Rebellion as the deadliest conflict in human history. The An Lushan Rebellion took place in China during the middle of the eight century and killed 36 million people, which, given that China’s total population at the time was 52 million, meant that two thirds of the population perished in the conflict. This immediately stank to high heaven to basically all experts on the subject, because contemporary accounts of the fighting don’t have anywhere near that intensity of violence described, and instead note that what’s actually likely to have happened is that the census got substantially less accurate after the rebellion: the figures come from the imperial census, which was conducted in 755 (before the rebellion) and 764 (after); the imperial state was also far weaker and much less capable of counting population. The idea that state capacity and institutional quality aren’t related is very weak: a 2009 paper by Timothy Besley and Torsten Persson concludes that policy choices are constrained by investment in legal and fiscal capacity, which are themselves determined by oolitical stability and good institutions - that is, they show that the quality of governance is not independent of the quality of institutions, which also trickles down to data quality: autocratic or backsliding governments, to take an obvious example, very frequently tampered with official data. And a 2021 study by Jonathan Hanson and Rachel Sigman develops a more in-depth relationship between extractive and coercive government capacity, and administrative capacity, that is, a relationship between the ability to keep the peace, raise taxes, and implement policy with a competent bureaucracy and a proper information infrastructure. The three dimensions are positively correlated, and administrative capacity is the dimension most closely related to conventional measures of institutional quality, even if they’re not necessarily a cause-and-effect thing.

There’s also a series of other problems, which mostly lie in the tension between AJR 2001 and the other landmark papers on institutions they published: a 2005 paper finds that Western Europe also had its development be related to the colonial Americas; a 2002 paper concludes that the richer a country was before colonization, the worse off it was after. The problem is that these two and AJR 2001 can’t all be true at the same time: if Europe only developed good institutions as a result of the interplay of its pre-existing institutions and American colonization, then how could it export them in a causally well-identified manner to its colonies? Again, there is a pretty vast literature on the subject of European institutions and colonial economic performance that complicates the picture significantly. And if pre-existing pre-colonial institutions are important, then how could they actually be explained by geography in the way the 2001 paper claims? A posterior study, by Stelios Michalopoulos and Elias Papaioannou, concludes that in the African example institutional quality was just straightforwardly caused by precolonial institutions only.

Economist John Holbein, in a sign of the times, used AI to do a quick fact check of Acemoglu’s work: what he finds is that it’s generally theoretically robust and empirically suffers from the critiques everyone points out, but it still stands after considering them. I do think it’s correct to raise questions about Acemoglu’s work and his methodological decisionmaking and to scrutinize the fact that one of the more famous people in the field is just doing shockingly shoddy stuff to prop up his theoretical conjectures. But everyone else’s papers are like that too. Nathan Nunn, the extremely well respected author of the paper I quoted about how the slave trade impacted Africa, has received an avalanche of criticism over the statistics of his paper (and the rest of his work on slavery) - in particular, the impact of the slave trade is only really present from 1970 to the year 2000, the one coincidentally discussed in the study. Another prominent paper by Oded Galor and Omer Ozak claims that time use preferences come from ancient agricultural practices; their evidence uses weights for land area that are inversed so the smallest countries have more impact, even though the paper states they don’t use those weights. Another two by Nico Voigtlander and Hans-Joachim Voth make serious econometrical errors of correcting certain geographical data sources in ways that, theoretically, they explicitly state can’t be don; all in a way that just so coincidentally strengthens their conclusions. Yet another extremely famous and well-cited paper by Melissa Dell, on the impact of Peruvian mining practices on economic development, also faced extensive criticism for its use of data and its also oh-so-convenient methodological choices.

Economist Andrew Gelman calls this the “Garden of Forking Paths”, in reference to a 1941 Jorge Luis Borges story of the same title. In the story, a man named Stephen Albert explains that a will left behind by the Chinese politician and scholar Tsui Pen saying he left behind “a novel, and a labyrinth” didn’t refer to two things, one of them an extremely confusing and chaotic book manuscript, but to one: the novel was the labyrinth, and it was so confusing because it showed every possible outcome of the same story. Tsui Pen calls this “the garden of forking paths”. For Gelman, the major problem in academia wasn’t “p-hacking” or “fishing”, where a researcher tests many theories and purposefully designs the final paper only in ways that heighten positive, easily published conclusions, but rather may inadvertently end up in the “Oh so convenient” stage because of constant choices around how to interpret the data, which data to include or exclude, how to transform that data into code, which transformations to apply, which controls to include, which model specifications to use, and how to handle data anomalies. In an extremely famous 1983 paper titled “Let's Take the Con out of Econometrics”, Edward Leamer showed that just by changing a handful of parameters in a model, you could get empirically robust results that the death penalty decreased, didn’t affect, or even increased the crime rate. Let’s face it, at the end of the day, we only care about tenure.

Theory or praxis

The perfect summary of the debate over Acemoglu’s work on institutions comes from a survey by the Institute for Replication conducted last year: among professional economists, they find that basically all of them agree with the theoretical case Acemoglu, Johnson, and Robinson make, but don’t agree with the empirical work.

But what is Acemoglu’s theoretical contribution? Basically, that institutions (the “rules of the game” of the economy) are important, and that there’s inclusive institutions, which are good, and extractive institutions, which are bad. In actual detail, inclusive institutions protect private property, uphold the rule of law, encourage innovation, and allow broad participation in economic and political life, whereas extractive institutions are designed by a small elite to extract wealth and power from the rest of society. Inclusive institutions suppress innovation and block long term development, while inclusive institutions permit sustainable prosperity. Let’s turn to The Economist:

…his grand thesis of institutions may not reveal very much. Nations prosper when institutions are good, and stagnate when they are bad. True. But what, exactly, is an institution? Rules, norms, enforcement, culture—everything, really. Where do institutions come from? From “critical junctures” and “institutional drift”, whatever that means. (…) Tyler Cowen of George Mason University noted that the institutional changes it describes only seem to come from other institutional changes, making the core argument regress for ever—turtles all the way down. Duncan Green of the London School of Economics has argued that the framework works mainly in hindsight. Francis Fukuyama of Stanford University has argued that the book waves away the example of China, which has seen blistering economic growth alongside institutions that can quite plausibly be described as extractive.

Again, nothing new, you’ve seen it here first. Let’s start with Fukuyama: for Acemoglu and Robinson (the authors of Why Nations Fail, the book in question), China is extractive, because their political system is highly centralized, unaccountable, and opaque. But, at the same time, its economic institutions are highly inclusive, because they focus on innovation and widespread wealth creation: for instance, between 2021 and 2025, Chinese retailers were forced to adopt a “refund-without-returns” model where a customer could request a refund and keep the item they purchased. They draw a historical parallel to the Republic of Venice, which had highly inclusive economic institutions that allowed for upward mobility through trade, while also being increasingly politically extractive via the Serrata del Maggior Consiglio starting in the 13th century, which banned new money families from participating in key government decisions; over time, this led to economic stagnation. Another parallel, for them, is the USSR, which also achieved extremely rapid economic growth under Stalin by shifting resources out of agriculture into industry; the soviets, according to Acemoglu and company, started stagnating once they needed to innovate on new technologies and new sectors. And that’s the objection Fukuyama raises: that China has managed to new technologies, like artificial intelligence, green energy, and electronics, even while also becoming increasingly authoritarian. Fukuyama’s claim is correct, though, while at the same time being kind of silly: the Soviet Union did stagnate and collapse just twenty years after beating the United States in all but one milestone of the space race. The idea that China being bound to decline over the long term is contradicted by China not declining over the short term is, well, to be blunt, stupid. Fukuyama is the one who oversimplifies the ideas in question. This applies to Green’s objection as well: basically anything can be explained in hindsight and not in foresight. American economists were predicting the USSR would overtake the United States in the late eighties. Are we all stupid now? Do we not understand how the social sciences work anymore? As Georg W.F. Hegel once wrote, “The Owl of Minerva takes flight only as the dusk begins to fall”. We’re still in broad fucking daylight.

I think Cowen has the most interesting objection: what isn’t explained by political institutions? Acemoglu is quoted as saying “…to understand institutional change from an institutional perspective you must consider the specific institutions in the past and the historical events…that influence them (…) I don’t think this is tautological at all.” Funny ending aside, he’s giving us a pretty insightful look into his thought process: in particular, he’s defending his institutionalist leanings from the perspective of the historicity of political life. Historicity is “the idea or fact that something has a historical origin and developed through history” (source: Wikipedia), and is one of the most important topics in continental philosophy, since it’s sort of the starting point for historicism, the “school of thought” (a school in the same sense Trump University was one) holding that all thought is conditioned by its historical situation, and that the historical situation itself is the variable of interest. The obvious problem here, one so glaring even Karl Slopper could point it out in The Poverty of Historicism, is that it’s impossible to justify historicism by its own standards. In the first chapter of History and Natural Right, Leo Strauss deals with the topic of historicism and the social sciences; in particular, he says, to be able to understand all of history through the vantage point of one particular moment, you need to assume that history has ended, in one way or another, at your specific worldview. Strauss emphasized that classical political philosophy (and the political philosophy that he believed ought to be done) focused on what the best regime (understood in more or less similar terms as Acemoglu understands institutions) ought to be, not on a value-free, incentive-based political “science” that described institutions without describing their purpose. The question of “inclusive” and “extractive” institutions is answered on purely instrumental grounds, through their attainment of economic prosperity. Whether liberal capitalist democracy itself is good is left completely unanswered. Well, sort of.

Political meh-ology

Acemoglu and some of his coauthors have two other books, which I, to be honest, haven’t read because they’re a thousand pages each. In The Narrow Corridor, Acemoglu attempts to give some sort of nuance to his rah-rah liberal institutionalism. In particular, he images two forces, state capacity and social mobilization, that leave a very narrow corridor for a country to end up prosperous and free. For state capacity, he uses the familiar image of the Leviathan, the biblical beast Thomas Hobbes used to describe the state; for social mobilization, an adjective. If the state is too weak, you get a deadbeat Leviathan that doesn’t enforce the law; if it’s too strong, you get a Nazi Leviathan that kills everyone; liberalism is the gorilla grip corridor where state capacity expands to meet new problems while civil society mobilizes to keep it in check. They call this the shackled leviathan (kinky) and state that it’s a stable equilibrium because of game theory mumbo jumbo.

Of course, aping the concept of the Leviathan is a bold move: Thomas Hobbes’s book is one of the most famous in all of Western thought and the foundation of political philosophy. Hobbes’s diagnosis starts in the English Civil War, a conflict so brutal it made him rethink the basis of human nature: for him, humanity started in a “state of nature” with no government, which, to him, was a war of all against all, where man was the wolf of man, and life was nasty, brutish, and short. Hobbes's argument is that only an effectively absolute sovereign can keep humanity out of the state of nature, because any division of power reintroduces the possibility of contested judgment about who has final authority, which itself produces civil war and violence. To go back to Strauss, this is the foundational act of modern political thought because it inverts the relationship between rights and duties: classical natural right starts from the duties owed toward realizing civic virtue and the common good. In contrast, Hobbes starts from a right, the right of self-preservation, which is absolute and equal for each individual because it’s the only right in the state of nature, and duties are derived from this primary right to sustain it. While most people think Hobbes and John Locke were at odds with each other, Locke is merely a successor of Hobbes: he believes the state of nature is a “good” one, where man is virtuous and rational, and merely voluntarily builds the state for his own self interest; in Locke, the right to self preservation expands to the right of comfortable self preservation, of owning property and increasing one’s wealth (which Strauss, banger factory that he was, calls the “joyless quest for joy”). Strauss thought Hobbes lowered the bar from virtue to survival; Locke lowered it even further, to comfort; Acemoglu, it seems, would lower it further still, to growth in GDP per capita and corruption in the Chart Crime Foundation for Lower Cigarrette Taxes rankings.

The other problem is the practical aspect: is it actually true that liberal democratic capitalism is self-sustaining? Take a look at a newspaper: no. Besides Strauss, there is one other major thinker working on the subject of Thomas Hobbes that, like a senile grandparent that keeps telling the same two anecdotes, this blog keeps trotting out: Carl Schmitt. Schmitt, for some relevant context, was a member of the Nazi party, served as kind of the consigliere of the early Third Reich, had some sort of fallout with Hitler during the early forties, and was spared from going to the big boy no shit you might get the death penalty section of the Nuremberg trials and was instead merely banned from teaching at a lesser hearing.

As a general thing, Schmitt believed that the way we understand “politics” (about taxes, regulations, transgender sports, etc) is pretty stupid and instead we should focus on Politics, which are about which people are enemies of the nation that have to die and which are its friends that are worth killing for. For Schmitt, this means the nation can come to a time of crisis, where the existing rules binding the state can't actually account for its capacity to resolve this unprecedented; comparing it to a biblical miracle, for Schmitt there has to be a sovereign underneath the norms who can decide, unilaterally and unencumbered, which of them to suspend. Hence, the famous opening to Political Theology, "Sovereign is he who decides the state of exception”. This means that the state has a value irreducible to the sum of individual interests, and that instead there is some sort of organic whole the Nation as a unit represents; and this faith has to be backed by some sort of real-world alignment of people and values. Liberal parliamentary government, in particular, was originally legitimated by a specific Enlightenment faith: that open, public discussion would arrive at the truth via the general will. By the 1920s, in contrast, mass democracy meant that parties represent fixed class and economic interests and decisions get made by political operatives in private arguments, meaning that liberalism had lost the social foundations that gave it its authority and legitimacy. And, in particular, the fact that liberalism gives a voice to all factional interest renders it completely incapable of decisive action in times of crisis; liberalism, rather than accepting that there is a possibility of genuine, irreconcilable conflict between these groups and the good of the whole, tries to dissolve their differences into other categories such as ethics and technocratic expertise.

There’s also, though, a second part. During his own time, Carl Schmitt had his own arch nemesis: the (non-Nazi) jurist Hans Kelsen, author of The Pure Theory of Law. In contrast to Schmitt, who argued that the survival of the body politic during national emergencies ought to be preserved by a single sovereign ratified by a plescbicite of sorts, Kelsen argued for a constitutional court: an impersonal judicial body applying written, deliberated norms. For Kelsen, the state simply consisted of the legal order, without any sociological or “spiritual” components. Even emergency powers had to emanate from written, formalized, interpretable norms - which even Schmitt new, since his early work for the Nazis consisted of justifying Hitler’s authority via Article 48 of the Weimar Constitution, which gave the Chancellor and President enormous autocratic power in case of a vaguely defined national emergency. But even then, Schmitt had a rejoinder: the exception of the titular state is definitionally the case the existing norms didn’t anticipate, and, even then, someone still has to decide whether this situation is the one the provisions cover. The process for determining and deciding the situations that count as emergencies, and the states of exception those emergencies permit, basically smuggled in the key decision of politics - a fact that ended up benefitting the Nazis, as the Weimar constitution was applied in an enormously lopsided manner to benefit the political right.

So the question, if we want to borrow Acemoglu’s terms, is how thin is the line between the shackled Leviathan and the absent Leviathan; his whole point is how liberalism cannot sustain human existence through any Leviathan. Instead, he turns to Hobbes’s final book, which details the “bad case” for government, titled Behemoth. In the Bible, the Leviathan is a beast of the sea; the Behemoth, in contrast, is the beast of the land. The Leviathan, according to Hobbes, describes the political commonwealth as an “artificial man” (Hobbes designed the book’s famous cover image), in which sovereignty is an “artificial soul”, government officials and laws are joints and nerves, peace is “health”, tumult is sickness, and civil war is, of course, death. To Schmitt, this was precisely the point: Hobbes created a “mortal god, to which we owe under the immortal God, our peace and defence”, who was some sort of Evangelion type robot.

In particular, this robot came with its own weak spot: Hobbes gave every man the right of self preservation, which included a right to an “inner life” (the right to refuse to act against one’s own preservation and thus freedom of conscience) which, in turn, opened a gap in the power of the absolute sovereign through which 19th and 20th century liberal constitutionalism crawled through and replaced the absolute power of the sovereign with neutral procedures, checks and balances, and legislative discussion. Much like Strauss, Schmitt would see Acemoglu’s worldview as a derivative of Hobbes’s, rather than as opposed to it; the shackled Leviathan is, when push comes to shove, no Leviathan at all.

Instead, Schmitt saw a superior alternative: the Behemoth. Schmitt defines the Behemoth as an irrational, chaotic, anarchic form of power, that embodies the “national spirit” and destroys the artificial, technocratic state through the power of myth and culture. To him, its victory over the Leviathan was inevitable: no man had any rights he was bound to respect; he could play as dirty as he wanted, and, thus, fight as hard as necessary to achieve victory. Of course, Schmitt’s prediction that this meant a victory for Nazi Germany over England and America did not come to pass, since this blog is written in English and not German. But why? I’m no history buff, but I assume the greater productive powers of the English and American states have a role to play; much like China and Russia, the advantage of the despotic Leviathan at dawn started to wane by noon.

Massachussetts: Institute or Technology

In their final book, Power and Progress, Acemoglu and Johnson (the Mike Wazowski of the three) put together a theory of technology and, well, power. This is their “Marxist” book. Noah Smith has a (negative) review, and Henry Farrell has a (negative) review of his review. It’s Emily Wilson versus Joyce Carol Oates for people with STATA licenses. Says the anti-Acemoglu article:

“Power and Progress” (…) treats a thousand years of innovation largely as a story of elite capture and unintended consequences. The cotton gin was the handmaiden of slavery, while the Haber-Bosch process created chemical weapons. Ordinary people apparently gained only when they could steer capitalists to consider their interests. The book plays down a much more compelling claim: that technological progress in and of itself has made the average person many times richer than in pre-industrial times.

The general thesis statement of the book is “Do not be fooled by the monumental technological achievements of humankind” Acemoglu and Johnson put together a 1,000 year history of the relationship between technology and “shared prosperity”, whatever that means because I guarantee it’s not really well defined, organized around the highly unoriginal thesis that technology can be good or bad depending on how it’s used. In their Nobel-winning academic work, the two (and Robinson and a few other guys who occasionally get invited to join the polycule) point out that the key difference between inclusive and extractive institutions is whether technological innovations can be blocked by a politically entrenched elites. I’ve used this in the past to explain how, well, Machiavellian the early political theory of AJR was. In the Discourses on Livy, Machiavelli argues that good institutions in Rome arose from the conflict between the masses and the elites over multiple centuries of (frequently violent) contestation and negotiation. Rome's liberty and its democratic institutions were a direct outcome of the efforts that plebeians made to express their discontent, a discontent that, to be fair, is never resolved: the plebeians want to be left alone and live freely, and the elites want to amass more and more wealth and power. Aristocracy always degenerates into oligarchy, which is then overthrown into democracy, which can then degenerate into anarchy; only by combining some democratic institutions with some aristocratic ones and some monarchic ones can democracy stabilize itself. Which is more or less basically the case the Narrow Corridor makes, with one exception: Machiavelli thinks political conflict is never solved, rather than being self-resolving through institutional channels over the long term.

Normally, Acemoglu and the gang assumed technology would be “good” and problems arose from good technologies being snuffed out by greedy aristocrats. Instead in Power and Progress they ask: what if the technology itself was the problem? A recent paper by Acemoglu and Pascual Restrepo, for instance, wonders whether automation benefits workers or not by (not) dissipating the wage premium certain jobs have. There’s nothing making it a law of nature that technological progress has to automatically translate into broadly shared prosperity, and whether it depends entirely on who controls the development of technology. During the Middle Ages, for instance, agricultural innovations enriched lords and immiserated peasants, while 20th century innovations matched with a specific institutional context (unions, government, and schooling) to result in rising standards of living. The new idea, it seems, is that institutions ought to be designed so that gains from new technologies like AI are captured by all of society instead of capital owners, who would benefit from mass displacement via automation and widespread worker surveillance. Ironically, Machiavelli might actually have something to say here: in The Prince, when talking about governent introducing new orders, he says “the innovator has for enemies all those who have done well under the old order, and lukewarm defenders in all those who might do well under the new”, which, first of all, sounds exactly like something Elon Musk would tweet, but second, is the entire problem of institutional capture: concentrated costs and diffused benefits. The people who introduce the new order are always a minority with clear interests, and the people the new order might serve are structurally at a disadvantage in resisting or redirecting it. One such example comes from a more obscure piece of Machiavelli’s writing: the Florentine Histories features an account of the Ciompi Revolt, a 1378 uprising by textile workers against guild masters who had monopolized control over the sector and were pocketing the political and economic profits from it. The workers took over the Florentine government briefly, and were then viciously crushed by the government; Machiavelli features one of the leaders of the Ciompi say, more or less explicitly, that power grows from the barrel of a gun.

You might notice that it’s not that hard to draw a line from Machiavelli to Carl Schmitt: once you open the door to the idea that the preservation of order requires force, then you need a strong government capable of enforcing said order, which then implies in certain cases the stakes are so high a completely despotic government is justifiable. Power and Progress is a deeply Schmittian book: technology’s trajectory is set by whoever holds enough power to impose their “vision” on everyone else, and their bet is that “the people” as a whole should hold that power in unity - but what happens, obviously, when different parts of a society have different opinions? Last year, for instance, the funniest economics paper since the one that examined whether trans people have a gender wage gap with themselves came out: after vapes are banned, queer teenagers are less likely to quit vaping. How can “the people” choose to steer the technology of vapes - to ease the mental suffering of closeted gay teens, or to protect their lungs from the ravages of cancer? The sovereign is he who decides the app of exception. The entire point of History and Natural Right was that, by closing the door on classical natural right, Machiavelli, Hobbes, and Locke had opened the door to Nazism; it is, for this reason, that Strauss described the Italian as akin to Columbus, discovering the continent where Hobbes “erected his structure”. If Strauss looked at the structure Acemoglu, Johnson, and Robinson put up further down the same stream, he’d have found three little piggies constructing a house made of straw.

And, of course, there comes another problem, and the one mentioned above: Acemoglu’s liberalism is predicated on a kind of materialism it cannot defend. Again, going back to Strauss, there is no clear argument why material gain, rather than some other conception of human flourishing, is the metric against which sixteen centuries of technological history should be judged - Acemoglu is, if anything too friendly to the idea technology-driven growth has been good for humanity. Ironically enough, the big villain of History and Natural Right, Martin Heidegger, has the perfect complement: Heidegger’s critique of technology (in The Question Concerning Technology) was that modern technology isn't a neutral tool we direct toward chosen ends, but rather a specific way of organizing human activities towards exploitation, optimization, and domination of “standing resources” rather than some conception of human flourishing. The danger, to Heidegger, wasn’t that any particular technology was bad, but rather that technological thinking had warped our relationship to the world so much that the furthest we can go is to ask “how can we use this technology to maximize material wellbeing”, which presupposes technological thinking.

A recent blog post by Nicholas Decker, recurrent Twitter main character, makes the eye-catching claim that “Daron Acemoglu is a Crypto-Marxist”: comparing his work, such as the paper coauthored with Restrepo, to old Marxist dogma around capital. The example is the Cambridge Capital Controversy, a debate between British Keynesians and American Keynesians over how to measure the stock of capital. The British side argued, persistently, that their model had to be more accurate because it was more realistic, and it also just so happened to have far more politically radical implications than the American model. Decker’s account is that the dispute was ideologically motivated (probably) and ideologically resolved (no); in fact, the main problem was that the American Keynesians couldn’t criticize the most obvious factor of the British models, that they did not describe reality with any semblance of accuracy, for the very simple reason that baseline Keynesian epistemology they both shared put an emphasis on the realism of the assumptions the model was based on over the realism of the conclusions it reached. Similarly, I think Acemoglu is reaching these “Marxist” conclusions (that technology exists to serve the class interests of capital owners regardless of the profitability of their endeavors) for a series of epistemological decisions that aren’t sustained anywhere: in particular, you cannot open the door to that kind of suspicion on technology without opening it on everythign else too. The philosopher Paul Ricoeur coined the term “the hermeneutics of suspicion” to described something common to Marx, Nietzsche, and Freud: reject the manifest presentation of a phenomenon and instead treat it as a symptom of a deeper, more real force producting it - class relationships, the will to power, or the unconscious. In particular, for Marx, pieces of “ideology” served not as themselves political maneuvers but as cover for the actual politics, which was class politics; behind anything, such as, for instance, “this technology is inevitable and you have to embrace it to not be left behind”, there is a class interest lurking, in this case, the interest of a narrow set of capitalist elites and their professional advocates in the economics profession. Rather than believing the participants’ account of their motives, Acemoglu says, look for what their speech is doing on behalf of whose power. But that’s where the buck stops - there’s never a concern over whether GDP or institutions or democracy or anything else is equally vitiated by capitalist interest. The people who take this furthest are the Frankfurt School, whose whole thing is that the moment The Odyssey started the world was doomed to suffer the horrors of the Third Reich. But Acemoglu has the good sense to not take himself there - which poses a problem: how can he say that a technology that seeks to optimize overall wellbeing (Larry Summers’s critique of his position on AI) is warped by capitalist ideology without actually unpacking whether his own life’s work is also just capitalist propaganda? GDP is fascist, AI is fascist, econometrics are fascist, manufacturing is fascist, and haven’t you read The Dialectic of Enlightenment, even going to the toilet is technically fascist.

Carl Schmitt called his book Political Theology because of his use of religious concepts to explain political events, particularly the idea of the state of exception as some sort of (evil) miracle. The problem with Acemoglu is he doesn’t have a political theology: he doesn’t take the leap of faith and say “this is what human life should be like”, which would, of course, consume his entire book; he dresses it up in some sort of technocratic patter and smuggles in the substantive concepts of that ideology via parliamentary and procedural rules, and through the very concept of institutions employed. For the longest time, the critiques of Acemoglu were primarily left wing: here you have an economist who thinks democracy can only justify itself via its preservation of property rights. Ironically his cleverest move was to try to beat Schmitt at his own game, by saying liberal democratic capitalism was a more stable arrangement than despotism; the problem is everything else is a bunch of legalese about the importance of the constitution. The post liberals are bringing fire and brimstone and you’re hitting back with the political implications of when the Talmud forbids selling donkeys after dusk. Without a substantive claim of what a good life is, Acemoglu cannot justify his skepticism of technologist ideology without also applying the same skepticism to everything else; if he tries to sidestep it with some sort of instrumentalism, he has no defense left over against totalitarianism.

Conclusion

Acemogly is frequently described as “Wikipedia with regressions”; a recent piece of commentary about him is that “unlike almost everyone else with those (technical) skills, he could also read encyclopedias and historiography handbooks, so to an ordinary economist c. 2008 he appeared almost impossibly erudite”. One of his recent papers, that I’ve discussed last week, is about the Marxist historian E.P. Thompson; if you find superficial readings of political philosophy plus shoddy econometrics unappealing, then why did you make it this far down this post.

But why did he get a Nobel Prize? Obviously because he’s famous and important and his papers drive a lot of debate. The intellectual history of development economics as a field is relevant. Originally development economists were mostly people who thought that investment in the right industries produced prosperity, which turned out not to be true and to not work. This followed a generalized reckoning issues in Keynesian economics that emphasized micro-level rationalization because government action could have unintended consequences. As a result, the new consensus was that free markets, free trade, and the free flow of capital was the right way to go. The problem here is that this approach did not work either: the nineties saw basically all “growth miracles” of the early decade collapse due to horrendous financial crashes in a globalized, interconnected world market. So if government intervention doesn’t work, and free markets don’t work, how can governments prevent catastrophic economic meltdowns?

This coincides with three very loosely related shifts in economics. The first is economic history becoming an actually rigorous discipline with solid empirical foundations that permitted credible work to advance really quickly. The second is economic history re-centering the notion of “institutions” through the work of Douglass North (and frequent coauthor Barry Weingast), which sought to answer the question of how contemporary government could provide both a stable, predictable climate for private investment, and also a credible and trustworthy government capable of providing social services by going back to the English Civil War, which, they conclude, created a system of government where the monarchy was bounded by clear rules - in particular, respecting private property. The leading practitioners of these two approaches received a joint Nobel Prize in Economics in 1993. At the same time, economists began coming up with new ways to prove causal relationships with slick and cool exploits that show “cleaner” relationships, which meant that, with then-cutting edge computer software, it became easy to prove how one thing caused another at the same time as troves of historical data was revealed. So, Acemoglu’s work came during a time with an intellectual climate hungry for answers about how to create a good government, troves of credible historical data, and a very refined understanding of how to use economic data. If someone with a Big Brain came along, he (or she, but almost always a he) put all the pieces together and answer, once and for all, why nations fail. Daron Acemoglu was that he.

The real question isn’t why he won the Nobel Prize: the research agenda of the last quarter century has been his research agenda. Isaac Newton once said he was “standing on the shoulders of giants” - shouldn’t the top awards in any field go to those giants, whether they eventually get slain by David or not? The real question is why anyone believed him, and the answer is that they needed to. To quote the great, perhaps even legendary, pseudoerasmus, who is among the people most genuinely knowledgeable about economic development and economic history, “AJR had such an influence, partly because there had been a perception circa 2000 that both state-led and market-led development had been disappointing. AJR’s use of history & colonialism rationalised this fatalism”. After the 2022 World Cup final, British football commentator Peter Drury said “Argentina, champions of the world. Again. At last. (…) 36 years ago since Maradona and Mexico, here finally is a nation’s new throng of immortals. Scaloni will be feted, Messi will be sainted. (…) He climbs into a galaxy of his own. He has his crowning moment…”. Economists, after three decades in the harshest wilderness, finally found their own Moses who could unite them and lead them home and resolve, at long last, the inquiry on the nature and causes of the wealth of nations.

If Strauss could compare Machiavelli to Columbus, and we can compare Acemoglu to Moses, then why not go even further and ask if he’s Odysseus: what if, like they ask in the movie, the moment he led his people home he realized the place he led them to no longer existed? That’s the merit he has over his critics: both of them have precisely the wrong answers to the major questions. But Acemoglu is the only one asking the right questions. He’s, alongside Fukuyama at times, the only one of the people involved who seems to realize how serious the situation is. Economist Luigi Zingales (who used to be a heterodox libertarian of sorts a few years back, but of course, who even is an orthodox libertarian anyways) compares Acemoglu’s opposition to technology to that of the Luddites; Zingales also, nonetheless, compares The Economist to Thomas Maitland, the man hired by industrialists to kill the workers.

When Acemoglu says The Economist wrote “a hit piece”, and accuses them of doing the bidding of Big AI in his (otherwise unconvincing and uninteresting response), he’s not entirely wrong. He’s also not entirely right. One of the major attacks on Acemoglu from the right is that his work “brings credit to bad political positions”, which is kind of implied by the original The Economist piece. Very recently, the president of the American Anthropological Association said she cancelled a panel on “whether biological sex still matters” because of its effect on political outcomes. How is that any different? Is this just yet another example of free speech hypocrisy by conservatives? Maybe. A better explanation, like most if not all academic disputes, is that this insufferable neurosis over who is allowed to speak at which places and/or collect which personal honors is that they’re all extremely full of themselves. Academics think academia is the single most important thing anywhere, which is why they treat us with endless blather about all their internecine office drama: they seem to be under the delusion that whoever gets the prestigious talks and awards gets to tell politicians and journalists and the public what to believe. They don’t. How many other times have you heard of the philosophy journal Hypatia? Zero.

What academics don’t know is that the garden of forking paths is the perfect metaphor for academic pursuits, because it’s a dark, bitter joke at the expense of the academics. In the story, Albert explains the garden to Yu Tsun, a descendant of Tsui Pen, and much more importantly, a Chinese spy for the German army who is fleeing British military police. After being found out, Yu Tsun escapes and takes the train to a random place he found in the phone book as a way to get his message out. Immediately after they’re done with the garden of forking paths, he shoots Stephen Albert in the head; his death makes the Germans know their target, a site with British troops and ammunition, is the village of Albert, in France. Whatever Tsui Pen intended for his garden to be, it didn’t matter, and there were bigger forces at play. Much like Stephen Albert, the critics of Acemoglu, and his defenders too, mistake how important their prattling about lacquer and ivory versus the man aiming a loaded gun at their heads. In Ancient Greece, there was a term used for someone who ignored the political context of their existence: an idiot. I will let the readers decide whether the shoe fits for any person involved in this discussion to wear it.