Kevin A. Bryan - Oct 13 2025
What a joy this morning, both personally and academically, to see a Nobel awarded for innovation. Joel Mokyr, an encyclopedia of economic history, was actually a PhD advisor of mine at Northwestern. Philippe Aghion and Peter Howitt wrote one of the three canonical models of innovation and growth that every economist in this area knows in their sleep (Loury and Romer being the other two). All three laureates are unbelievably nice, positive, friendly colleagues, in a way I find it hard to overstate. We could fill this article just with stories - bagatelles, shall we say, for Philippe's sake - but let's intersperse the fun with the rigor. As I know Joel so well, I'll devote most of this essay to discussing his work and the intellectual history behind it, before looking at Aghion and Howitt's incredibly influential Schumpeterian model. To understand the most important event in economic history, the Industrial Revolution, or the most important future event, the impact of AI, it's hard to think of a better place to start than the work of these three.
To understand where Mokyr, Aghion and Howitt are coming from, we must start with a stylized history of economics research. Early economics - think Adam Smith and Ricardo - was about cross-sectional differences in the level of prosperity across nations. Subsequently, the core question of economics became how to prevent recessions. This was unsurprising, given the Panics of the 1870s and 1890s, the Great Depression, and the related wars. But since sometime between 1957 and 1962, the primary question of economics has been about growth.
Heidi Williams and I looked at why this shift happened a few years ago. Schumpeter was of course already well known by the middle of the century, and indeed while at Harvard was the highest paid academic economist in the world. He believed that innovation and entrepreneurship (not invention, per se) were economic phenomena responsive to incentives. Schumpeter was not atheoretical; he was just bad at it! As Samuelson explains: "If it was not given to Moses to enter into the Promised Land, it is understandable that he should have tended to exhalt the promise of that fair country. So with Schumpeter...It was typical of Schumpeter's love for theory that he rejected Marshall's view that the reader could skip the footnotes and appendixes. If time were short, Schumpeter advised, read them and skip the text!"
But despite Schumpeter's interest, the field at large did not consider innovation to be a first-order economic phenomenon - we knew that new things happened, but how and why seemed to be less important than how new factories got built. However, Bob Solow's thunderbolt 1957 decomposition suggested that economic growth could not be explained by more people or more capital per person, but rather by a "residual", a dark matter many suspected was the creation and diffusion of new ideas. Zvi Griliches' thesis work, also published in 1957, looked at the diffusion of ideas using a formal empirical model, in his case to see why corn hybrids took so long to be adopted across regions. Innovation as a key economic driver was in the air.
Inspired by that work, the Rate and Direction NBER conference, with a published 1962 volume, began investigating these ideas formally. The volume included Arrow's famous study of why innovation may be hard to incentivize. By the mid-1960s, we had formal government statistics measuring aspects of R&D, especially via the OECD Frascati manual. We were seeing heavy government support for aspects of innovation following Vannevar Bush's "Endless Frontier" call to continue the benefits of World War II research. And economics journals were becoming filled with research linking the microeconomics of research to the macroeconomics of growth, both in more orthodox spheres and among Nelson and Winter's evolutionary models, Freeman's Innovation Systems work at Sussex, and Nate Rosenberg's technological history. As Lucas said, "Once you start thinking about growth, it's hard to think about anything else," but the coda should have been "and growth means innovation".
This was the world Mokyr, Aghion, and Howitt entered as they pursued their PhDs in the 1970s and 1980s.
Let us begin with Mokyr, who comes into innovation via the side door, or indeed the chimney. His Yale dissertation was not about innovation, or even the Industrial Revolution, but rather about the Low Countries, economic powerhouse of the 17th century, that of tulips and the Dutch East India Company and Vermeer. Mokyr was born in the Netherlands just after the Holocaust, and though he grew up in Israel, the question of why Holland didn't continue as the world power must have seemed natural. But how to answer? Mokyr was a student of Bill Parker, the cliometrician and editor of the Journal of Economic History in the 1960s, driving the idea that economic theory plus statistical analysis could answer historical questions better than a deep reading of a dusty tome in the stacks (with the caveat that at Northwestern, if one's fingers didn't get dusty at some point, you were considered a bit of a poseur when it came to economic history). At the time, this was called "The New Economic History".
Joel's early work continued to triangulate around England using this New Economic History. First the Low Countries, then Ireland, largely due to his longstanding research relationship with the great Irish economic historian Cormac Ó Gráda. Ireland was poor pre-famine, among the poorest places in Europe. But why? More importantly, how shall we know? Mokyr: Clarkson (1980) noted with some relief that present work in Irish economic history was "a trifle old fashioned, with the bewitching voices of the social sciences muted and statistical wizardry missing. Consequently Irish economic history is generally readable, if not always profound. Wizardry or not, economic history without the social sciences and statistics is becoming increasingly unthinkable...[A] list of possible explanations of Ireland's poverty can be made, based on observations of contemporaries, on theories proposed by historians, or on models which are based on modern economic analysis and social thought...as a rule a deliberate and explicit attempt will be made to confront the argument with data..."
This was a method many objected to. Indeed, Mokyr's book, Why Ireland Starved, says surprisingly little about the Famine itself! The New Economic History looked where economic theory and data could shine a light, period. And there was no period, in the other sense, more important to shine a light on than the Industrial Revolution. What could matter more in history? We went from a world where the richest places and times were maybe five times subsistence, where in the 1840s there was a famine in the United Kingdom that killed a greater percentage of Irish than Pol Pot did Cambodians, where there was a common job for six year olds to climb into small mine shafts to clear out dirt - we went from that world to the modern one. Was it institutions? The English patent system? Colonialism? The "wave of gadgets swept over England," as Ashton's student said?
Giving some formal evidence as to what happened in the Industrial Revolution was something of a holy grail, and one that many of the best economic historians of the past generation - Allen, Mokyr, Crafts, Harley, de Vries, McCloskey - worked on. But as Mokyr mentioned in his first essay on the topic, "[t]he New Economic History, with its instinct to treat complex historical problems in a formal, deductive fashion, has thus far not attempted to encroach upon the Industrial Revolution. The reasons for this omission are obvious enough: The bewildering variety and paralyzing complexity of the interaction of economic, geographic, political, and social factors would render any attempt to model the Industrial Revolution as a whole a futile exercise in pretentiousness."
Pretentious exercise or not, it was worth the try. We can already see Mokyr's later theses in this essay. First, he notices a pattern based on his incredible knowledge ("Joel has read everything" is a good stylized fact). In this case, he notices that a lot of early modern manufacturing in Europe appears where there is a pre-existing rural home production, or small factory production, of related pre-modern goods - such as linens produced at home versus in a factory. Why the correlation and what does it mean? Let's write a set of production functions. The rural economy consists of farmers who can do home production or grow crops, choosing how much of each to produce to maximize earnings. A new technology arrives using capital to produce goods. These new goods compete with home production, so you can attract workers who were doing light industry at home. Profits are plowed into factory growth, pulling more folks in. High agricultural productivity regions therefore industrialize later.
Fine as a starting point, but we can push further. With a model like this, how and why will the factory improve its productivity and output? Mokyr was visiting Stanford early in his career, where Arrow was writing his learning-by-doing models, and Paul David applying them to innovation. As so we see in Joel's next Industrial Revolution paper arguments less from pure history than from the economics of innovation: what types of shifts in demand lead to more spending on innovation? Here we find the debates of Elizabeth Gilboy, Schmookler, Rosenberg - debates on what must be true as a matter of economic theory for changes in factor prices or quantities to lead to improvements in productivity, and what data could settle the debate.
As Mokyr worked through various potential explanations for the Industrial Revolution - colonialism was not more extractive, demand did not induce innovation supply, there was Protestant Work Ethic elsewhere, the level of science was not particularly world-beating - he arrives at an idea that is novel: "upper tail human capital", the 5-10% of folks between the social elites and the common worker. Not elite scientists inventing new ideas, which France and Germany and Song China and ar-Rashid's Baghdad and Salamanca all had. And not general education. In the late 1700s, most people could still not read in England. But England uniquely had a sufficiently large base of people educated enough to understand and diffuse new things, a culture and incentives that pushed those folks into industry, and an openness to ideas that come from people who are not social elites.
This idea has an antecedent that is often missed via the ideas of Mokyr's longtime colleague Jonathan (Jon) R.T. Hughes.
Hughes is credited often by Mokyr, but few economists have likely read his magnum opus, The Vital Few. Hughes was a football-playing man of the West - born in Twin Falls, Idaho, schooled in Utah, but brought into the profession via a Rhodes Scholarship and a PhD at Oxford. He spent most of his career at Northwestern, with a stop at Purdue alongside Nate Rosenberg and the early applied game theorists, including his frequent collaborator Stan Reiter. Reiter would later join Northwestern's MEDS in the era of Holmstrom, Milgrom, Myerson, Stokey, Satterthwaite, and Roberts. Purdue and then Northwestern were wild places in that era, largely on the basis of hiring the wackiest, most ambitious folks you could find, conditional on them knowing economic theory: for instance, Vernon Smith, the Nobel winning experimental economist, was also at Purdue at the time. Personally, as a Mokyr student who formally studied at MEDS, and who was, to my knowledge, the only one to have ever had both Satterthwaite and Mokyr on a PhD committee, I found it interesting that modern economic history and modern economic theory were already linked by the early 1960s via Hughes and Reiter! No surprise this is Hughes: "I think modern history will force economic historians to ask questions which are not data-based in nature, but are certainly to be explained by economic theory."
Hughes is out in Idaho around the time when cities are just being built out of scrubland, when folks like Philo Farnsworth are inventing the television. While historical tomes emphasize national leaders or "social forces" as drivers, how can one not notice in a small town that the handful of people who pool funds to open a factory, the rector of the agricultural extension college, or the town booster are what really matter? And that these folks, the Mary Switzers who largely built the modern American social welfare system, or the Andrew Carnegies who built modern American industry, are not just a bunch of wealthy dilettantes who think in grand halls, but people who do things? This was Hughes' point in The Vital Few.
Mokyr was looking at biographies and histories from 18th and 19th century England where, boy, you see the same pattern. The Lunar Society, the Society of Arts, and the SDUK bring information to smaller towns. Folks like Watt, son of a shipwright, and Arkwright, a barber, and Trevithick, son of a mine foreman, play really important roles in the development of new technology. Of Mokyr's four "big books" on the Industrial Revolution - Lever of Riches, Gifts of Athena, The Enlightened Economy, and A Culture of Growth - I would recommend Gifts. The argument is simple. To get the Industrial Revolution, you need enough growth, fast enough, to outrun factors that push back against growth like Malthusian mechanisms or institutional constraints. This means new ideas must arrive and diffuse quickly. To do this, you need enough of an epistemic base, in this case the ideas of the Enlightenment, plus a culture and mentality of experimentation, plus a society that is willing to and able to apply these ideas to economic practice. England's Upper Tail Human Capital, culture of "tinkering", sufficient base of science, and industrial focus allowed what would have been infrequent scientific advances to become frequent advances, embodied in products, and spread to industry. The knowledge existed, could be improved, and was made useful. Ergo, growth.
I'll reserve further detail here - Anton Howes, among others, have nicely discussed the evidence we have on Mokyr's thesis. See also my earlier writings on Mokyr's theory and MBA-level slides on the Industrial Revolution. If you want one article-length summary of Mokyr's views, I recommend either his Editor's Introduction or his "Precocious Albion" chapter with Kelly and Ó Gráda. I also love his edited volume with Baumol and Landes called "The Invention of Enterprise", looking at what different historical societies pushed their upper tail human capital to work on. As for other explanations of the IR, my notes above give a brief introduction, but Pseudoerasmus has a good post on Allen's coal thesis in particular. Had Mokyr not won with Aghion and Howitt, a joint prize with Allen would have been reasonable.
Let us now turn more briefly to Aghion and Howitt. Howitt, trained in Canada and long a professor at Brown, has been instrumental in establishing that institution as a center of research on long-run growth. Aghion is both an incredible growth theorist, but also an applied theorist with much broader contributions. By my count, he has 27 papers with over 1,000 citations, from institutions to growth to organizational theory (Formal and Real Authority, with Tirole, is one of the great applied theory papers of all time). No surprise he is one of the legendary Collège de France's two economists, alongside Esther Duflo. There was a conference last year for us innovation economists at the Collège - by far the fanciest place I've ever gotten to speak! - in honor of Griliches and featuring a keynote by Mokyr. Many of the subfield's legends came, and you realize just how much intellectual firepower was brought to bear on understanding growth between 1960 and 2000.
Recall our history of growth theory in the profession. Following Solow (1957), we had to understand what exactly the dark matter that led to growth was, if it wasn't just capital deepening. That is, we had to determine how innovation mattered for growth and what the incentives to create and diffuse it were. Paul Romer in the 1980s, and then Chad Jones, figured out the technical apparatus called "endogenous growth theory" necessary to include ideas in a production function such that we didn't just get exploding growth or growth that petered out. As Grossman and Helpman pointed out, earlier neoclassical growth models seemed counterfactual both because growth eventually died out and also because technology was exogenous - dropped by God to be used by firms. But the many case studies of highly-innovative industries, historical and modern-day, showed active investment and spillovers from that investment were common. Hence endogenous theories - those where innovation was endogenous to purposeful activity by firms - was necessary. But initial endogenous models missed something critical both to Schumpeter and to histories going back to the Industrial Revolution, or even earlier: what does growth do to incumbents?
In "A Model of Growth Through Creative Destruction", Aghion and Howitt, just a few years after Aghion got his PhD, present a "Schumpeterian" endogenous growth model. Glenn Loury (an early colleague of Mokyr's at Northwestern) had a decade earlier proposed a seminal model of patent races, where firms invest to try to improve a technology and thus seize a market. Schumpeterian models attempt to put this idea into general equilibrium, and into a dynamic setting with new inventions following old. Incumbent firms are either leaders or laggards in an industry. Leaders try to pull away technologically while laggards try to catch up. A counterintuitive but natural result can be derived: in a very competitive innovation market, rents are destroyed quickly and hence innovation is unprofitable, while in a very uncompetitive one, the leader is so far ahead of the laggards that it isn't worth it for the latter to try to catch up or the leader to try to pull further ahead. An intermediate level of competition is best. Excessive competition in the innovation sector, due to the ability of laggards to "business steal" by catching up with innovation leaders, can create too much incentive to innovate, not too little.
The trick here, of course, is how you can model this battle between business stealing and competition escaping behavior in a tractable way - Aghion and Howitt's setup is particularly elegant. Grossman and Helpman's quality ladder paper, also Nobel-quality, builds on this idea in a way that is now widely used to study the effect of innovation policy. In both, the key insight goes back to the '62 Rate and Direction volume we mentioned early in this essay, via Ken Arrow's famous "replacement effect": incumbents have less incentive to innovate because innovations "steal business" from their own current products, while laggards have less incentive to innovate because the market they can apply the innovation to is quite small until they catch up to the frontier.
An intriguing tension exists here, on two grounds. Are new inventions complements or substitutes for incumbents? And do new inventions reduce costs, giving competitive markets the "biggest market" for inventors due to higher quantity produced, or raise quality, creating the opposite market structure link? Any combination of the four leads to quite different predictions. For instance, Netflix is a complement to telecommunications infrastructure providers, not a substitute. AT&T and Deutsche Telekom want Netflix to innovate - no business stealing at all. What one might want as a model to guide innovation policy, in the broadest possible extension of Aghion and Howitt, is one that can handle underappropriation of the rents from innovation across each of these four possible cases; don't worry, we're working on it!
Are these Schumpeterian models empirically useful? Absolutely. For instance, ignoring the incentive for incumbents to innovate to escape competition, one would expect monopolists to innovate more, as their profits are higher from the new good. Empirically, however, we do not see this relationship. Indeed, Aghion, Howitt and coauthors have a famous "inverted U" paper showing that, as theory predicts, intermediate levels of competition have the highest level of R&D. There are many critiques of this particular result, but the general finding where monopolists are not the R&D kings does seem to hold across many cuts of the data. In a beautiful study by Aghion, Acemoglu, Bursztyn and Hemous, optimal innovation policy in a Schumpeterian model combined with climate modeling shows the benefit of first subsidizing research until clean energy is price competitive with dirty energy, then taxing the latter only at the point (closer to the policy the world is actually following than Kyoto-type interventions)
The history of the Industrial Revolution and the nature of Schumpeterian growth are also critical for the future of AI. Both empirically and theoretically, the early decades of the Industrial Revolution were not particularly good for wage-earners. Joan Robinson, as quoted in Mokyr (1976), noted the worrisome "low level bastard golden age" where wages could have fallen to subsistence with mechanized Industrial Revolution production. So will we get the low level bastard golden age, or the utopia, or something in between?
Taking Mokyr seriously, the "Culture of Growth" - that is, the embodiment of AI inventions into processes, laws, and organizations - matters more than the inventions themselves. One can easily imagine situations where AI could contribute to growth but does not due to frictions similar to those in the early Industrial Revolution. France and Germany did not experience delayed growth because they were too unintelligent to see what happened in England, but because their societies lacked the tinkering mentality to bring those ideas into their industrial systems. The second big barrier for AI also comes from Aghion himself, in a paper written with Chad Jones and Ben Jones (another advisor of mine!) for the inaugural NBER AI conference here in Toronto in 2017. Here, they allow AI to automate nearly all industrial processes. This leads only to finite, constrained growth because, due to Baumol effects, the remaining unautomated industries increasingly constrain future growth. Intuitively, consider farming: as agricultural productivity took the labor force from 50% farmers to 2%, the impact of further automation on unprocessed food production becomes tiny.
