The Institutions Without: Daron Acemoglu, The Economist, and Who Gets to Steer Progress

In October 2024, I reflected on the work that brought Daron Acemoglu, Simon Johnson and James Robinson the 2024 Nobel Prize in Economics. Their central insight was that prosperity depends on the architecture of institutions: who has power, how widely opportunity is distributed, and whether a system can correct itself before its weaknesses become permanent. I extended the idea into organizations and personal life. We are all, in some measure, institutional architects.

That essay later became part of my collection, The Human Centre, a series of reflections concerned with a simple question: how does one remain real inside systems that reward the appearance of reality? Institutions occupy a central place in that inquiry because they can preserve judgement, agency and character, or gradually replace them with credentials, metrics, visibility and performance.

The past fortnight has supplied a rather less abstract sequel. One of the world’s most influential institutions of public argument has placed one of the world’s most influential economists on trial.

On August 17th, The Economist published an unsigned article titled “The world’s most influential economist is oddly unconvincing”. It opened with the observation that, after a few drinks, unnamed economists sometimes reveal their true opinions about Acemoglu. Apparently, peer review now comes with a drinks menu.

The article ranged across three decades of scholarship: the disputed historical data behind Acemoglu’s work on institutions, the breadth of the categories used in Why Nations Fail, a recent paper on falling birth rates, and his sceptical estimates of artificial intelligence’s effect on productivity. It then converted these disagreements into a judgment about Acemoglu himself: perhaps his reputation at the summit of economics exceeds the quality of his work.

Acemoglu replied sharply in a letter published by the magazine. He welcomed serious criticism, accused the article of assembling loosely connected grievances into an insinuation of unreliable scholarship, and returned the magazine’s verdict to its sender. His response was convincing. A credible defence of Acemoglu begins with his fallibility.

The legitimate argument

The empirical dispute surrounding Acemoglu, Johnson and Robinson’s famous 2001 paper on the colonial origins of development is real. The authors used the mortality rates faced by Europeans in former colonies as an instrument for the kinds of institutions colonial powers established. Where settlement was dangerous, Europeans tended to construct extractive systems; where settlement was feasible, they were more likely to establish institutions protecting property and encouraging investment. Those institutional differences then persisted.

David Albouy’s 2012 critique identified serious problems. Of the 64 countries in the original sample, 36 had been assigned mortality rates derived from other countries. The underlying sources mixed soldiers in barracks, soldiers on campaign, labourers and other populations whose experiences were not necessarily comparable. Once Albouy corrected for these issues, the relationship became weaker and the instrumental-variable estimates much less reliable.

Acemoglu, Johnson and Robinson responded that Albouy had removed much of Latin America and Africa, relied on inconsistently coded variables and produced results highly sensitive to particular choices. The debate remains unsettled. A 2026 survey of 352 experts found no consensus, with opinion leaning slightly towards Albouy on the empirical dispute while retaining broad agreement that institutions influence development.

There are conceptual objections as well. “Institutions” can encompass laws, norms, enforcement, political culture and historical inheritance. A category that broad can become better at explaining the past than predicting the future. China’s long period of extraordinary growth under plainly authoritarian institutions remains a difficult case. Institutional change can also appear to arise from earlier institutions, producing Tyler Cowen’s memorable image of turtles all the way down.

Acemoglu’s recent work on AI carries another vulnerability. His 2024 paper, The Simple Macroeconomics of AI, estimated fairly modest productivity gains over the following decade. He has since acknowledged that the model did not anticipate the rapid development of agentic AI and gave insufficient attention to possible gains in scientific discovery. In a field moving faster than the usual publication cycle, a two-year-old estimate can acquire the atmosphere of an Assyrian tablet.

These are substantial questions. They fail to sustain a verdict on a 33-year body of work. A scholar as prolific as Acemoglu creates an unusually large surface area for criticism. Selecting several contested claims from hundreds of papers tells us the size of the available dossier. Serious evaluation requires a denominator, comparison with his peers, and some account of what his research programme allowed the profession to see that it had previously neglected.

Economics is deliberately adversarial. Seminars, referees and replications exist to expose weaknesses. John List, an economist who has collaborated with Acemoglu, made the crucial point in his response to the article: counting objections reveals very little about the ultimate value of a research programme. Every ambitious body of work develops amendments, counterexamples and failed estimates. A pristine corpus usually belongs to somebody who has published very little.

The Economist had ample material for a serious assessment. It chose a different genre: the reputational prosecution assembled in the style of a pseudo-literature review.

The verdict in search of evidence

The handling of Acemoglu’s recent demographic paper is particularly instructive. The Economist cited a social-media observation by Jesús Fernández-Villaverde that estimates drawn from historical declines in fertility may fail to extrapolate to a future of much lower birth rates. Acemoglu’s paper makes precisely that point in its conclusion. The magazine acknowledged the caveat, then suggested that such qualifications become lost beside his “reams of maths.” Fernández-Villaverde, according to Acemoglu, was never interviewed.

Far from being an oversight, including this limitation highlights a commitment to rigorous, carefully measured scholarship.A model describes a defined world, identifies relationships within it and states where extrapolation becomes hazardous. The article transformed that boundary into evidence of shallowness. It thereby rewarded the posture of certainty and penalized the disclosure of uncertainty.

This is an odd standard for a publication that has spent decades warning governments, businesses and investors against excessive confidence. The measured scholar begins to look evasive beside the pundit who has never met a confidence interval.

The article also criticized Acemoglu’s institutional theory for explaining too much in hindsight. Its own method was strikingly retrospective. The headline supplied the conclusion; the writers then travelled backwards through hundreds of papers, gathering whatever disputes could be arranged beneath it. The procedure resembles the weakness it attributes to Acemoglu: a framework spacious enough to absorb every available fact.

“Oddly unconvincing” is itself a category of remarkable elasticity. It accommodates contested data, acknowledged caveats, disagreements over China, uncertain AI forecasts, anonymous professional jealousy and one blogger’s decade-long irritation. Acemoglu is criticized for defining “institutions” too broadly by a publication using “unconvincing” to cover an entire intellectual life.

The Economist borrowed the authority of academic criticism while dispensing with its more inconvenient disciplines: representative evidence, consistent criteria, attributable claims and serious engagement with the author’s replies. Its editors have perfected a technology for sounding like the consensus before the consensus exists.

The anonymous oracle

The Economist has long published most articles without bylines so that its writers can speak in a single institutional voice. The tradition has produced clarity and consistency. Here it produced a peculiar asymmetry: an unnamed writer, backed by an authoritative global brand, relied partly on unnamed economists to deliver a highly personal judgment about a named scholar.

Academic peer review also uses anonymity. Its anonymity is embedded in a process. A referee must engage the paper, an editor assesses the report, and the author receives an opportunity to respond. The argument leaves an auditable trail even when the referee’s identity remains concealed. The Economist’s version supplied the prestige of a committee and the accountability of a rumour.

Esoteric vocabulary offers an unexpectedly useful metaphor for this arrangement. An egregore is a collective form sustained by shared attention, belief and repeated ritual. In modern institutional life it appears as culture, consensus, brand, mythology, incentive and fear. It eventually seems to possess a personality independent of the individuals who created it.

The Economist’s anonymous house voice resembles an institutional egregore with a red masthead. Individual writers dissolve into a single personality that surveys governments, markets and civilizations from a position of cultivated altitude. At its best, this collective identity preserves memory and editorial standards. Under less favourable conditions, it begins to defend its own reflection, interpret dissent as deficiency, and mistake a familiar tone for demonstrated truth.

This is one of the concerns I explore in my forthcoming book, The Age of Narcissus. Narcissism begins as misrecognition: the reflection is mistaken for reality. As the mirror expands, narcissism becomes architecture. Organizations build brands, hierarchies, incentives and rituals around a preferred image of themselves. They can eventually admire their own seriousness and mistake that admiration for evidence.

Reality-based governance requires clear claims, honest tests, durable memory, correction without shame and incentives that reward accuracy over performance. Image-based governance relies on narrative, status and the maintenance of the institutional mask. The Economist article belongs uncomfortably close to the second category. Its prose sounds authoritative because The Economist sounds authoritative. The brand becomes a form of evidence about the quality of its own judgment.

An institution convinced of its seriousness can become unserious without changing its voice.

The magazine eventually published Acemoglu’s reply. This deserves acknowledgement. Allowing the subject to answer after the reputational headline has travelled around the world still resembles minimum due process more closely than editorial magnanimity. A publication that routinely lectures nations about institutional accountability might have examined the institutional design of its own prosecution.

A civil war within liberalism

Acemoglu remains a liberal. His new book, What Happened to Liberal Democracy?, proposes what he calls “working-class liberalism”: shared prosperity, stronger communities, pluralism, political voice and technologies that increase the value of human expertise. Philosophically, it draws upon freedom as non-domination. The older liberal reflex concentrates on coercion by the state. Acemoglu’s field of vision includes private actors whose command over employment, data, platforms and political access can become equally arbitrary.

This creates friction with a market liberalism that treats technological progress as a largely undifferentiated good. The Economist argues that technological development has made the average person many times richer than people were before industrialization. The historical observation is broadly true. Acemoglu and Johnson’s Power and Progress asks how long those gains took to reach ordinary people, which groups captured them first, and which institutions forced a wider distribution.

The disagreement concerns the politics inside the machine. Technological invention expands what a society can produce. Ownership, bargaining power, labour organization, regulation and democratic pressure determine who receives the gains and whose way of life becomes expendable along the way. The steam engine did not write the Factory Acts.

One anonymous economist told The Economist that Acemoglu uses sophisticated models to justify populist policies that have already failed. “Populist” has become a wonderfully efficient word in elite policy debate. It can now describe almost anyone who mentions workers before the final paragraph.

Acemoglu’s recent offence against the liberal household is straightforward. He has begun treating concentrated private power as a threat to freedom. He regards technology’s direction as a political-economic choice. He wants workers represented within that choice. These propositions remain entirely compatible with a serious liberal tradition of pluralism and non-domination. They create considerable discomfort for a narrower liberalism that has grown accustomed to scrutinizing the state while treating the corporation as part of the scenery.

The timing deepened the impression of ideological boundary maintenance. The article appeared alongside a dismissive review of Acemoglu’s new book. The review framed the book primarily around AI even though, as Acemoglu observed, AI and democracy occupy only about half a chapter. He interpreted this as an effort to obscure the book’s larger message and return the discussion to The Economist’s favoured defence of rapid AI development. The timing supplies circumstantial evidence of a coordinated editorial frame; motive remains an inference.

Editorial priors, status systems and boundaries of respectable thought can produce the same outcome without an elaborate conspiracy. The collective voice can enforce them without a meeting ever being called. That is what makes an egregore a more useful metaphor than a smoke-filled room, although The Economist has already supplied the drinks.

Who already governs AI?

The Economist raises one genuinely important objection to pro-worker AI. Acemoglu and other economists have urged society to steer AI towards complementing people. The magazine asks who this “we” might be and who should decide which technologies qualify.

Any adequate answer begins with the people making those decisions today: research laboratories, executive teams, corporate boards, venture funds, patent owners and the small number of firms controlling advanced models and computing infrastructure. They choose which technical possibilities receive capital, which products are built, and which measures of success govern their deployment. Consumers arrive much later, selecting among the paths that survived the investment committee.

Market choice begins downstream of capital allocation. Tax codes may make machinery cheaper than labour. Publicly funded research supplies foundational discoveries. Intellectual-property rules establish which returns can be privately captured. Government procurement creates early markets. Data law defines the raw material available to model builders. Even the supposedly unregulated path has been engineered through institutions.

Treating prices as neutral spirits speaking through a séance conceals the structure of ownership, law and bargaining power through which they are formed. Prices communicate information inside that structure; they cannot confer legitimacy upon it.

The status quo is a governance regime. It assigns control rights primarily to capital owners. Steering begins with capital allocation; public policy can broaden the hands on the wheel through worker participation, revised tax incentives, public research, procurement standards, competition policy and rights over workplace data.

Acemoglu’s long work on directed technological change gives this argument greater substance than The Economist allows. Innovation can follow trajectories that replace labour, create new human tasks or make existing expertise more productive. Early choices generate skills, infrastructure, business models and political constituencies that reinforce the chosen path. By the time a failure becomes obvious, the alternative ecosystem may no longer exist.

Recent scholarship also gives “pro-worker AI” more content. In Building Pro-Worker Artificial Intelligence, Acemoglu, David Autor and Simon Johnson propose technologies that help teachers, nurses and skilled tradespeople make better decisions and acquire expertise. Separate 2026 work by Ajay Agrawal, John McHale and Alexander Oettl formally models AI that raises worker productivity without automating the worker’s task. The idea may sound obvious. Building an economic system that rewards it is considerably less so.

The Economist treats democratic steering as a dangerous intervention into technological progress. Corporate steering receives the gentler description of innovation. The distinction says more about the magazine’s theory of legitimate power than it does about technology.

Productivity, power and the human centre

Suppose Acemoglu has substantially underestimated agentic AI. His quantitative forecast will require revision. The political economy questions grow more urgent as the technology becomes more capable.

More powerful systems can raise productivity while concentrating ownership, reducing labour’s bargaining power and weakening the communities organized around work. Acemoglu’s newer research explores these secondary effects. AI, Human Cognition and Knowledge Collapse models the possibility that reliance on generative systems will weaken incentives for human learning and gradually degrade the knowledge on which the systems themselves depend. Automation and Repression examines how automation, falling labour shares and inequality may alter a state’s choice between redistribution and political control.

These papers are models. Their contribution lies in widening the set of outcomes placed under examination. Aggregate productivity captures only one dimension of the transformation.

We might also speak of democratic productivity: a society’s capacity to generate independent judgment, preserve plural sources of knowledge, distribute bargaining power and correct collective mistakes. An economy can produce more measured output while impoverishing those capacities. A firm can become more efficient while rendering its workers more dependent. A state can acquire extraordinary technical competence while becoming less accountable.

This is where Acemoglu’s argument meets the human centre. Human beings bring judgement, agency, tacit knowledge, responsibility and character to a system. A technology that preserves and enlarges those capacities contributes to human development as well as output. A technology that renders them economically superfluous may still produce an impressive quarterly number. The long institutional ledger may record a different result.

Agentic AI therefore creates a double uncertainty. Its economic benefits may be much greater than Acemoglu first estimated. Its institutional consequences may also be deeper. Confidence about the first supplies no automatic answer to the second.

Technological systems also create mirrors. They tell institutions what can be measured, predicted and optimized, then encourage those institutions to reorganize themselves around the reflection. The danger described in The Age of Narcissus appears here in technological form: the model becomes more legible than the human reality it represents, and the institution gradually mistakes legibility for truth.

A human-centred economy keeps the relationship in order. Models remain tools. Institutions retain memory. People preserve the authority to judge circumstances that fall outside the metric. Efficiency remains answerable to human purposes.

The institution in the mirror

In “The Institutions Within,” I argued that healthy organizations allow ideas to circulate, develop talent and preserve the long game. Leaders create extractive environments when they hoard authority, suppress challenge and pursue immediate gains at the expense of institutional vitality.

The present dispute brings that framework back to the institution producing the criticism. A publication that judges scholars is itself a governance structure. It determines which ideas receive attention, what qualifies as evidence, whose voices are named and how readily a subject can answer.

In its treatment of Acemoglu, The Economist concentrated authority in its collective brand, drew personal testimony from the shadows, and presented ordinary scientific contestation as evidence that the object of its attention had somehow escaped scrutiny. The episode reveals the strange power of institutional prestige: a publication can borrow credibility from its history while questioning whether a scholar has borrowed too much credibility from his.

The mirror expands. An article ostensibly examining Acemoglu becomes an examination of the institution that commissioned it. The Economist set out to ask whether Acemoglu’s reputation justifies his influence. Its chosen method raises a parallel question about whether the magazine’s reputation justifies the authority with which it speaks.

Acemoglu can be wrong about settler mortality, China, demographic extrapolation and the speed of AI. His larger contribution remains formidable. He has placed institutions, power and the direction of technology at the centre of economic inquiry. His present insistence that liberalism rediscover the working class has disturbed a comfortable settlement between professional elites, capital owners and technological optimism. That disturbance deserves a response at the level of argument.

The Economist supplied a charge sheet, anonymous witnesses and an atmosphere of settled judgment. Its evidence showed something considerably more ordinary: an important economist has produced ambitious work, encountered criticism, acknowledged uncertainty, defended contested choices and revised some expectations. That record offers no proportionate basis for the prosecutorial structure. Science has survived worse.

Liberalism’s survival will depend on extending its scrutiny of power to corporations, platforms, technology owners and its own educated establishment. Acemoglu offers one possible route. His programme requires criticism, empirical testing and institutional experimentation. His critics require standards equal to the authority they claim.

A society that treats the current ownership of technological power as part of nature has already surrendered the constitutional question. The invisible hand in this debate has a board of directors, a cap table and government contracts. The Economist has mistaken its grip on the wheel for freedom.