Hello, welcome to the Innovation and Diffusion podcast. I'm Ruveyda Gozen, a research economist at the London School of Economics at the Program on Innovation and Diffusion.
And I'm John Van Reenen, and I'm a professor here at the London School of Economics, and we are delighted to have our special guest this week, Daron Acemoglu, who is amongst many other things a professor at MIT, an Institute Professor at MIT, which is the highest level of professorship you can have. And you know, Daron and I used to be colleagues at MIT, and I miss that every day. He's one of the most interesting, dynamic thinkers around in economics. I remember once having a talk with Daron and others, and we were thinking about how to be, when we think about promoting people or how are people successful in economics, and Daron said, well, you have to own an area, there has to be an area which is yours. And then somebody said, I don't know if it was me or somebody else, said, but you know, Daron owns like 10 areas. And you know, if you look at his career, he has worked in so many areas: technology and the labor market, which we'll talk about, political economy and politics, long-run institutions, what makes nations successful or fail, and most recently in his book, which we'll talk about, with Simon Johnson, Power and Progress: How Vision, Ideology, and Politics Shape the Direction of Technical Change. So really, welcome here.
Thank you. It's a great pleasure to be here on the Innovation and Diffusion podcast. And John hid the fact that before we were colleagues, we've also been co-authors, so you know, there's some history there.
Yeah, I'm also very excited to have you here. I mean, I'm really excited about this podcast and episode actually. So, and today, I mean, we're going to go into your book, your Power and Progress, but I also want to ask you things about, you know, who Daron is, why he is interested in economics, and you know, and his like production function and those kinds of things. So I want to start with the question, you know, what made you interested in economics? Were you interested in politics and economics, you know, when you were a teenager? You know, did you have a role model that made you follow this path?
You know, if this was, if this podcast was about one of my former books, previous books with Jim Robinson, Why Nations Fail, this would fit even more because actually I got interested in economics when I was a teenager in high school. And the things that drew me to economics were exactly the topics that then became, you know, the centerpiece of Why Nations Fail: institutions, democracy, you know, why are people opposed to economic growth and cause poverty, you know, triggered by the context of Turkey. You know, this was in the mid-1980s when Turkey was a military dictatorship, chaotic, very high unemployment, and little did I know, I thought this is what economists studied. So I said, I want to, I want to do economics. And then my father said, if you go to a university in Turkey, you'll end up in jail with your outspokenness and your views, you know. So he said, you have to go abroad. I'm not going to come and pick you up from jail. So I started like a mad dash to find somewhere that would take me abroad, and I ended up at the University of York, and actually studying economics and politics.
I think geography has some effect on our choices. I think if you look at most of the Turkish students' statements of purpose, for instance, they would write, oh, I've been through, you know, many like chaotic things in Turkey, like, you know, the hyperinflation and all kinds of problems, and that makes really people interested even more in economics. And as you mentioned, you've written a lot of books. Some would say too many.
Yes, yes, a lot of articles as well, we should say.
We should put a tax on Daron because, you know, he makes the rest of us look bad as he produces. So the most recent book is called Power and Progress. Well, I mean, well, I don't know if it's going to sound weird, this question, but why did you write this book? Like, what made you write this book?
You know, this sort of evolved from my understanding, however imperfect it may be, that there is more to both the distributional aspects and also the importance of the choices that we make on technology than I had realized. That, say, in the 2000s, you know, I wrote that one of the books that perhaps John has in mind, I wrote a graduate economic growth textbook, and there's nothing on these issues in that book. That book was sort of parting the conventional wisdom of, you know, technology endogenous or it's my own work, direct the technology, but the full implications, as I see them now, of how different types of technologies create different distributional effects and the preconditions for productivity gains, all of these things were missing in that book. And this is what I started realizing in the 2000s, you know, around 2010, and then that was sort of eye-opening for me, and it pushed my research in this direction. But the even more immediate trigger for writing the book is that once I started looking at the world through those lenses of, you know, automation both creating a very clear set of forces that redistribute income between capital and labor, between different types of labor, and not always by itself being a recipe for shared prosperity, that perspective then coupled with the techno-optimism or the obsession with AI that had emerged in the United States sort of made me think this was an important topic to tackle. You know, fortunately or unfortunately, since then I think there has been more of a questioning attitude towards AI in the US sort of press as well. But you know, when we started, when I started thinking about these things, optimism was the only game in town when it came to technology.
Yeah, some people are also on the other hand very pessimistic, right? That it's, I mean, there are two edges in this picture. Like one side is very optimistic about the future, and then the other side is very pessimistic. So in that sense, maybe this book provides both sides. And then what we find, I mean, empirically and, you know, on technology and its effects on society. So, and also, like, I mean, one of the main topics, like hot topics in economics right now, I mean, we know that productivity has been declining, labor share is also declining, that you also mentioned that in your papers. So what role does automation play in this trend?
Well, you know, let me actually say one thing before I answer your question, which actually gets to the heart of it. You know, there is a lot of alarmism about AI and technology, but I actually think, I wouldn't say we are somewhere in the middle between the optimists and the pessimists, because I think that actually the optimists and the pessimists are equally wrong, and I don't want to be in the middle of them. So the pessimists are actually not pessimists. So they are completely optimistic about the capabilities of AI, and they just worry about things like, oh, you know, is it going to create superintelligence and how we're going to control it. And I think that's like swallowing the line from Silicon Valley hook, line, and sinker. It's the wrong kind of pessimism. You don't need super capabilities, which I am very, very suspicious of, by the way, in order for these technologies to do a lot of damage. A really bad scenario is what we call so-so automation, meaning automation that's very pervasive but not super productive and controlled by a very narrow elite, together with information controlled by a narrow elite. That's not the scenario that most of the pessimists are worried about. So I think they actually distract from the real issues, in my opinion. There are other people, they're friends of mine who say I should all support, you know, the alarmism because you need a big coalition. But, you know, that may well be true. But I, for the record, I think that's the wrong kind of pessimism.
Just to, you know, elaborate a little bit for the listeners on the so-so technology, which, you know, Daron has, you know, pioneered the thinking about. These are technologies which are in a sense directed against groups of workers, say less educated workers or workers in general, and actually don't create much benefits or relatively small productivity benefits. Daron sometimes uses the example of, you know, self-service supermarket checkouts, which is something which replaces or will replace a lot of checkout people, but, you know, actually probably creates a lot more cost for the customer, right? I mean, you know, many, I mean, there are many different aspects of so-so technology.
I'll come back to your question about, you know, the effects of automation. I think that's very interesting, but just, you know, there are many, many types of automation technologies that at the end of the day disappoint in terms of their productivity. Some of it is by design because they are actually meant as redistribution technologies. They really take away rents from one group to another. Others, they really shift costs to customers, you know, self-checkout kiosks or customer service. Others just like, you know, oh yeah, there's a little bit of profits for shareholders, but the damage in terms of redistribution is much greater. So there's a whole spectrum of these so-so technologies, and it all goes to the question of what has automation done over the last, you know, 40 years. And, you know, I am absolutely not against automation. I think automation has always been a central part of the industrial process ever since the beginning of the British Industrial Revolution, sometime around the 18th century, and it's going to be in the future. And if we use it right, it's actually very useful because, you know, if you look at some of the tasks that have been automated, they are the really unpleasant, very heavy ones, and like factories are much safer places today because of robots. But a singular focus on automation, especially coupled with so-so automation, is damaging, and I think the last 40 years shows that. So my research with Pascual Restrepo, and a lot of the research is with Pascual, so I'm really grateful for that collaboration, my research with Pascual Restrepo shows that, you know, estimates that in the United States about 70% of between-group inequality, inequality between detailed demographic groups such as those with high, low education, gender, age, etc., about 70% of that increase in inequality is related to the use of automation technologies. And while those technologies have been super important for changing the landscape of who gains and who wins, their aggregate effects on TFP, total factor productivity, economists' staple measure, has been minimal, very, very small. So they have not led to a productivity revolution. So a lot of it looks like, you know, some, you can find examples, robots and car factories have been great, but on the whole, these technologies have not led to a productivity revolution. And I find it very difficult to believe that amazingly disappointing productivity experience we've had over the last 40 years, despite the enthusiasm about computers, is partly because of this misdirection of technology.
Yeah, I think, I mean, most of the time we tend to think that, you know, technological progress is a wave that will lift all the boats in the society. But then, I mean, based on your research, we see that actually not everyone benefits from technology.
Actually, yeah, I mean, I think again, you know, you have to distinguish different versions of it if you want to get a bit of a more nuanced idea. I think everybody, even the most techno-optimistic people, would accept that there is misuse of technology, like nuclear technology. You can use for nuclear energy or you can use it for making nuclear weapons. And even people like Marc Andreessen and Peter Thiel, who, you know, are amazingly enthusiastic about anything that new technology will do, I'm sure even they wouldn't defend nuclear weapons. But, you know, I remain ready to be surprised. So, but once you leave that, you know, very obviously damaging uses of technologies, there is a view both in the tech circles and in economics that goes like this, which is, sure, some people will win, some more than others, but if new technology expands our capabilities, that means we can produce more, we can do more things, and ultimately that's going to create a very powerful set of forces that are going to be beneficial for everybody. And what are those very powerful set of forces? It's going to lead to better, cheaper products. It's going to lead to better services like healthcare and education, but also critically, it will lead to higher real wages. So real wages will increase both because we get cheaper products and also because firms are going to be more tempted to go and hire workers because they have become more productive. And as a result, you know, they will bid up wages. And most of us earn our living in the labor market. So, you know, that's the channel via which that rising tide is going to lift most boats at least. And, you know, even before, you know, I started doing research in this area, it was clear that that story didn't completely hold together because in the United States, everybody knows median wages have been, you know, almost stagnant, and there has been a huge increase in inequality. So it looks a more nuanced picture at the very least.
But I mean, that view is just, it's kind of conventional, like I guess probably closer to my conventional view, the sense that, you know, technology creates the, expands the size of the economic pie. Now, nothing tells us that's going to go to the median worker because the political and social institutions might mean those benefits get channeled to the top 1% or the top 0.1%.
So that's kind of, I mean, more like the kind, maybe conventional sense. I guess the right view would be will definitely trickle down. The more center-left view is, well, you need to do things right to make sure it does trickle down. But Daron's view is a bit more radical than that, which is he'll explain, it's like the technology itself is almost hardwired, right, not to give the benefits to the average worker, right?
I wouldn't say hardwired, but there are paths, yes, there are paths of technologies. I think the critical thing here is choice, that there are many paths of technology, and there are paths of that technology which are going to be very, very adverse for the working class. And I think even the sort of the center-left view, I think here the center-left view between the US and the UK may be a little different. In the US, you know, most of the center-left would also believe that there are strong forces for wages to increase out of technological progress, but they may think, well, inequality may increase, and you have to do something about inequality, or you may need to shore up institutions in some ways because there will be some people who won't adapt, and they will be unemployed. You know, after all, if there's one thing that's hardwired, it's this channel from technological improvements to higher wages, and it's hardwired into economics thinking, and economics is actually quite influential in these debates in the United States. So in sort of emphasizing that actually you can have technological progress and that could reduce wages, you get a lot of pushback, let me put it that way.
I mean, you know, so the economics, so it is actually interesting and simple, though it may be, it's useful to reiterate, you know, when we say technology improves our capabilities, a summary way of stating that is that technology is going to increase average labor product, which means that we produce more with the amount of labor that we have. And what the presumption is in economics and in a lot of discussion is that if average product goes up, wages are going to go up. But actually, when you start thinking about it for a second, you'll realize that in economics, that's not implied. In economics, wages are in competitive labor markets equal to, in other settings generally proportional to the marginal product of workers, what the workers bring to the production process. And the reason why there is such optimism about technology in economics is because we have built into our models direct and indirect mechanisms via which average and marginal productivity move together.
And I think that's completely unwarranted, that's my view, that there is no reason, and history is very consistent with this, that average and marginal productivity co-moving is something that could happen for some technologies and not for others, and that's where also institutions and other arrangements can matter. And this is such a simple idea, but it isn't part of the canon, and it's not part of the general discussions. So that's why I feel compelled to sort of repeat it. And if you want an example, the one that we give in the book is the sort of oft-quoted story of the future factory, which will have two employees, a man and a dog. The man is there to feed the dog, and the dog is there so that the man doesn't touch the equipment. So if that is the future of the factory, and some people think that's a great future because, you know, going back to 1950, many management journals were sort of saying, look, workerless factories, that's great. So I don't think it's a science fiction that we are heading that way. If that's the factory, we have huge average product, one employee or two employees if you count the employees, if you count the dog, and a huge amount of output, and we're going to keep on increasing with better equipment, that average productivity is not going to translate into higher wages because the humor of the story is that that man is not really that useful. So that wedge between average and marginal product, I think that's critical, and you cannot deny it.
And I mean, I want to ask a few things. One is about history. The other one is that I was thinking right now, so there should be some things, you know, since we also talked about power and how technology is somehow, you know, driven in a certain direction. So what makes these people to, you know, if it's a so-so automation, what makes them choose automation over labor? Are there things that make it more, you know, attractive, you know, to switch to capital instead of, you know, labor?
I think that's a great, great question, and that's a part of the book, so that's why we had to write a book, not just a few articles, which I have done. I think there are institutional, power-related, and economic reasons for over-automation, and I think you want to delineate those, and because that then sort of provides a much better understanding of what the alternatives are. And I think the most important three reasons I would emphasize for over-automation, and I think each one of them is somewhat controversial even within academic circles. One is that managers always have a bias. Its strength may vary from institutional setting to institutional setting, or from organization to organization, but managers have always a bias for control. And what that means is that the more you depend on labor as an essential factor, the less you control. So that's going to imply that managers are always going to want to automate more than what may be socially desirable in order to claw back that control. So in an economic model, that would, for example, one way that that would exhibit itself is workers get rents, for example, because of bargaining or efficiency wages. Great papers by John Van Reenen showing that, oh, my Google Scholar citation rate is going to the roof. But if those are those rents, if I automate as a firm, I claw back that rent, those rents. But if I was a social planner, I don't care about them. So that's one aspect of that control. Second, now going from the less controversial towards more controversial, I think the general attitudes of managers and stock markets and society, for example, as bracketed by the rise of mega companies and the Friedman doctrine of the social responsibilities of business is just to increase profits and give a higher rate of return to shareholders. All of that says it's great to cut labor costs, and because labor costs are a very important part, and if you cut labor costs, at least in the short run, you can deliver a better rate of return to shareholders. So that really pushes companies towards automation. So that's a kind of an ideology come economic force. But actually, third, and this occupies a good chunk of the later quarter of the book, there's also an ideological bias in the tech industry, which I trace to Turing's brilliant work, which was mathematically brilliant but philosophically flawed, and the whole AI field that got shaped by that, which is that it elevated autonomous machine intelligence to become the ultimate aim of technological development. And if you think that the way you should judge progress is by reaching human parity, developing more and more machines that act, sound, and perform tasks like humans, that creates an immediate bias for automation. So it's this bias coming from the corporate sector together with the bias coming from the tech sector that really sort of created this excessive push towards automation. And you see what excessive push for automation does is really pernicious if you have my framework, because automation is necessary, automation is great. The first tasks you're going to automate are going to be the ones that are really socially very useful to automate. You get big productivity gains, they are helpful, helping in terms of eliminating these very difficult tasks. But if then ideology pushes you into more and more automation, you're going to get into these marginal tasks that are not really bringing any productivity gain. So you get all of the disruption, distributional negative consequences, but you don't get any of the productivity benefits of automation. So this excessive automation is what the book is warning against.
So on that third point, the kind of the ideology of the Turing trap, AI, the Turing trap, I mean, were, I mean, I should remember this from the book, but were the philosophers, scientists articulating an alternative vision?
Yeah, so I mean, I think we're in a better position because there are alternatives, we argue, both technologically, and these are not green fields that have never been tried. They have been tried and much more successfully, I would say. So right away, there was a very different vision of what automation was about that came from MIT's own Norbert Wiener, the famous Wiener process guy and cybernetics guy. And he had, you know, as early as 1950, he was seeing what robots would do, and he was trying to articulate a set of ideas how, you know, computers can work better with humans. And I think it had some influence in terms of both some students following him and in terms of the philosophy. So the cybernetics books that he wrote became quite well known in some philosophy circles, but Wiener himself was not like a hands-on engineer. But then a few other hands-on engineers picked this agenda up. I think the most two notable ones are J.C.R. Licklider, who became like the grandfather of human-machine complementarity, and even more importantly, Douglas Engelbart, who was a very hands-on engineer, but also he developed a very clear articulation of what we should have. So Douglas Engelbart, for example, was himself invented the mouse, hyperlinks, hypertext. His students and he together invented a lot of the things that you're using on your cell phone in terms of like menu-driven systems. And all of this was coming from a philosophical and applied agenda which said, you know, we can make computers much more usable for humans, increase human productivity, increase human capabilities. And the way that we articulate that in the book is we give this a name, we say machine usefulness rather than machine intelligence. But second, we also try to give the theoretical foundations against, again, drawing my work with Pascual Restrepo. The most important thing, and this is where the choice and direction of technology issues become really central, is not like making workers more productive in the task that they perform. That can help a little, but that's never going to be powerful enough to counterbalance automation. And we argue both in history and in the future, the more powerful and actually positive path both for distribution of resources in society and productivity is creating new tasks. So when you create new capabilities for workers, more complex tasks, more tasks that are involved in new types of work, new types of product, those are the ones that really bring human creativity and human skills centrally into the production process, and they create this uplift for labor share and new jobs and wage growth. And machine usefulness is really at the center of it. And the other reason why we wrote the book now, well, or perhaps we were just lucky, coincided that generative AI actually has great capabilities for both automation and for creating new tasks. So the choices that we are talking about, and they're technological as well as institutional, are even more important today in our current generation than they may have been, you know, in the past.
And also, I mean, since we talked about the creation of new tasks, I think education also plays an important role there, right? Because now, I mean, we're talking about automation, and then we need to learn more and more. I mean, if there will be new tasks, maybe, I mean, maybe, I mean, some of those tasks will not require like a high education level, but then we should also invest a lot more in education then, right?
Well, you know, thank you for bringing that up. And at the expense of making myself even more unpopular with economists, I'll actually give a more nuanced answer to that excellent question. So yes, you're absolutely right, education is critical because if you have new tasks, you need new skills. That being said, you will see that we don't mention education in the book many times. There's a very good reason for that, because I am very dismayed by the following line of argument among economists, which I think has been part of the problem, which is, oh, you know, if technology is going in a direction that's not good for some workers, the solution is education. To everything, the solution is education. And, you know, that is both, I think, the wrong solution, and it also cements the wrong way of thinking, which is extremely common. And this is part of the institutional response that, you know, the left, the center-left articulates, which I object to, which is, well, let technology go where it is, where it will go. We can't control that. We shouldn't control that. The market knows best. And then if the technology creates a mismatch with the skills of the labor force, the labor force should get more skills. And if you're unsure, let's send more people to college. And I think that has not been a good strategy, and it's not the right way, because first of all, the technology has many, many different directions it can go. There is no guarantee that it will, the market process will guide it to the right place. Second, if there is a possibility of mismatch, most of the time it's better to meet somewhere in between rather than let technology go wherever it is and people try to catch up with it. Third, there is a lot of evidence that just trying to sort of create this two-tier structure where you send as many people to college as you can, but then those who are left behind are even in a worse position because you've created this two-tier structure of good skills versus bad skills. I think that's been extremely pernicious in the US and the UK. So I think the overall way in which economists have thought of education, especially higher education, as the Swiss Army knife solution to all problems, I think has been deficient. We discussed this extremely briefly in one paragraph.
But I'm sort of now opening it up. So I'm also very interested in economic history as well. So, and in your book as well, I mean, you go into the first Industrial Revolution, you talk about, you know, how the economy has changed. So one of the things, so maybe we can talk a little bit about, you know, the relationship between social change and technological change. Do you think that they, I mean, feed to each other or one predominantly determines the other? You know, I mean, when you think about the first Industrial Revolution, like before that, I mean, we have aristocrats and then, you know, landlords and then, and then we have a new, like a different society and society is also changing. What do you think about this relationship between social change and the technological change?
Absolutely, the two go hand in hand, and what is cause, what is effect, not clear. You know, Simon and I are not economic historians of the British Industrial Revolution, but we think that era is very, very important to understand our current...
I mean, you have quite a deep discussion of it in the book.
Yes, well, we pretend to be, but...
I was going to say six different explanations to the Industrial Revolution discussed.
Right, right, but we do two little, little, tiny contributions to that debate. One is we actually propose a somewhat different than the established six different explanations, which is very much in line with what you just said, social change preceding the Industrial Revolution. This is the sort of the middling sort of people who come from not the lowest background, but not from the aristocracy or the gentry, who become empowered because of a long process of institutional change, and who become sort of motivated by ways of trying to both enrich themselves and change the world via new scientific ideas and new social ideas. And they are at the heart of the British Industrial Revolution, I think that's what we sort of argue. And in fact, we think, you know, that's much more likely to be the distinguishing factor between, say, Britain and France or other areas, that social change was stronger in Britain. But the second is that my younger self would have looked at IRA and would have said, look, this is great, this is a new class of people rising, piercing through the existing hierarchy. That's fantastic. You know, Britain is turning from a very hierarchical, top-down, semi-aristocratic society to one in which, you know, people who come from modest backgrounds are becoming rich, are becoming powerful. But there is a more nuance to my thinking on that. In fact, there is a bit of a parallel between that middling sort of entrepreneurs, like people like Richard Arkwright, and the Silicon Valley, that they're both piercing the existing hierarchy, but they are also very elitist in their vision, and they want to recreate a hierarchy with themselves at the top rather than create a more egalitarian system. So if you look at the more powerful middle class, say, very beginning of the 19th century or the end of the 18th century, it's extremely condescending towards the lower classes, even more so than the aristocrats. And it's no accident that, you know, this is the period in which things like poor laws become much harsher. And it is exactly the same thing that Silicon Valley is piercing certain hierarchies, but their vision is so anti-worker. You know, workers are these fallible, dispensable people. The less we have, the less of these humans we have, the better. So there is a bit of a parallel between those two things, and that's like the nuance that I like to emphasize.
I would love to talk a little bit more on the historical aspect of things, but due to our time constraint, I also want to talk a little bit about the relationship between democracy and automation. So you talk about surveillance systems and automation and how it affects, you know, the democratic regimes. So what is your opinion on that? I mean, about the relationship between automation and democracy, like how, what is the power of government and how do they use that power against people maybe or in favor of people maybe?
Yeah, so I mean, again, I think those are really, really central questions to me, and we may have slightly simplified it in the book. I think it's a very complex set of relations, but I see automation and how you use digital technologies in an automation-centered way as inherently anti-democratic. There are at least three senses in which that's true, but again, it's more nuanced, I accept that, but these three pathways from this particular approach to automation and technology to democracy are: first, if automation creates much more inequality, especially with lower class, I mean, lower education workers losing out, that's going to be very, very inimical to democracy. You cannot, I am very convinced, you cannot have democracy in a society in which 50, 40% of people are left behind. So I don't think in the current system in the United States, for example, we can have a well-functioning democracy. I do not find the rise of Trump and all sorts of crazy ideas both on the left and the right as surprising against this background. Control of information is critical to democracy. You cannot have democracy in a society in which the information is controlled by a handful of actors. So you cannot have democracy in China of any form until the monopoly over information of the Chinese Communist Party is broken. And likewise, I don't think you can have democracy well-functioning if all information is controlled by Google, OpenAI, and a few other companies like that. And then the third, there is an anti-democratic ethos to automation, especially as articulated by, you know, some Silicon Valley leaders, because it sort of starts from the premise that most people don't know what's good for them, they're not really good for their jobs, you know, you need to cajole them and you need to limit their influence both economically and socially. And I think that's again very difficult to make work with democracy.