 Welcome to InetLive. I'm Bill Jane White, co-founder of the Institute for New Economic Thinking and author of Doing Capitalism in the Innovation Economy. This episode of InetLive is on the future of work in the context of the long-term digitalization of economic life and under the immediate impact of the COVID-19 pandemic. I'm here today with Daron Asamoglu, Killian Professor of Economics at MIT. And given that productivity is a central subject of our conversation, I should add that Professor Asamoglu was arguably the most productive economist alive in the world today. We will discuss meaningful integration or jobless future, and we'll examine how we can shift the odds towards a future of inclusive prosperity with meaningful integration between humans and digital technologies. At the end of our discussion, I'll moderate a Q&A from exports and the audience. Please submit your questions anytime by clicking on the Q&A button. Now let's get started. Daron, you have published a number of articles addressing exactly this issue of automation in jobs, robots in jobs, the wrong kind of artificial intelligence. You've been a leader in the academic examination, empirically and theoretically, of how technological innovation impacts labor markets, occupation, skills, and earnings. Perhaps you might summarize for us the way that you distinguish the two impacts, the displacement of jobs by the impact of automating technology, on the one hand, and the reinstatement of jobs on the other, and the way those patterns have evolved in the post-war era. Thank you, Bill. It's a great pleasure to be here, and thank you for having me on for discussing these issues, which are very close to my heart and very important. Let me start by saying that technology always and everywhere has major distributional consequences. It's not just about robots or automation or AI. It's always been like that. And we have different types of technologies, often pushing in different directions. So the way that I would like to think about the period of rapid and shared growth, say from the end of World War II to the late 1970s, is not that we had technologies that were automatically making everybody better off. It is that we had an institutional framework, and within that institutional framework, different types of technologies were being introduced and rolled out in a way that counterbalanced each other. So there was plenty of automation, and automation always by substituting capital for labor makes the production process less labor intensive, displaces workers from certain tasks, but at the same time, there were other economic opportunities spearheaded by new firms, by new technologies that created gainful employment for all kinds of workers. So at the heart of this dynamic is exactly the pair of concepts that you mentioned, displacement and reinstatement. So we think of automation as displacing workers from the tasks they used to perform, because now they can be performed by machines or algorithms. But at the same time, new organizations, new technologies, new products create new tasks that are labor intensive, that reinstate labor centrally into the production process. The best example of that, perhaps, is what happened during the mechanization of agriculture at the beginning of the 20th century and at the end of the 19th century, when obviously the big breakthroughs displaced workers in large numbers from agriculture, that's very well understood, new jobs for these workers weren't created automatically. They were created because firms made efforts to find new technologies in industry, in found factory floors. For example, you have a number of new industries spearheaded employment growth. You have, for instance, the rise of much more rationally organized factories with non-production workers, design workers, engineer, supervisory workers, back office workers. And you have, of course, the expansion of the service sector. And all of these were critical for both U.S. growth and the fact that it was basically generating jobs, enough jobs for people to be able to find ways of sharing in that prosperity. Your work and others have shown that that worked pretty well in the United States from the end of World War II through about 1980. But since then, since 1980, there's been a real shift, a discernible shift, in both the, well, in three factors. One, slowdown in productivity growth. Two, more displacement than reinstatement. And three, among workers, a substantial increase in inequality from the top of the labor market towards the middle and bottom of the labor market. Yes, you summarized it perfectly, Bill. Thank you. So I would put the more displacement than reinstatement at the center. That's really very important. And you can see that, for example, from the fact that you have a huge decline, starting in sometimes in the mid 1980s, huge decline in the labor share of manufacturing. But no other sector is increasing its labor share to counterbalance that. And then my own work shows that this was one of the most important factors in the rise in inequality. Workers who were in jobs, routine jobs, production jobs, some service jobs that were being automated during that period were at the bottom end and were falling further and further behind. And the puzzle is, which I think nobody fully understands, is during this period, we also had a productivity slowdown. And that's really remarkable, especially from the 1990s onwards. We are going through one of the periods of recent history where you have the largest number of innovations. The number of patents is skyrocketing, especially digital patents. The share of electronics and computer patents is rising from something like 20% to 60%. And it's just new gadgets everywhere. And despite that, we're not seeing productivity gains. Now, my suspicion, and there's some circumstantial evidence to back this up, but we don't know for sure, is that that is also very closely related in that it is connected to the fact that we are not adopting a balanced portfolio of technologies. So we are going too much in the direction of automation, too much following a particular vision, running into diminishing returns and not exploiting other very fruitful directions. Lance Taylor's new book that was sponsored by Inet calls out the phenomenon of the high productivity sector shedding labor, displacing labor, and labor being absorbed in the lowest skilled, lowest paid, lowest productivity occupations. The classic example might be the machine tool operator in the factory that gets closed who winds up eventually getting a job in a McDonald's or a Walmart or as an orderly in a hospital. And that shift in the pattern of employment across sector may be playing a role here. Do you think? Absolutely, absolutely. 100% you know, what my work with David Otter and David Otter's work with David Dorn, for example, shows is that many of these automation technologies impacted occupations and jobs in the middle of the income distribution. And moreover, these were jobs in the middle of the income distribution that were very critical for, you know, for lack of a better term, I'll use it as American dream, meaning workers who didn't come from very well off backgrounds often did not have very good education, may have a high school education or two or three years of college or some specialized education in the process, but then could get into relatively high wage manufacturing jobs or jobs in some specialized service industries that was very important for building a middle class existence. And these are exactly the jobs that automation targeted, both because they were expensive jobs so you could do a lot of cost savings with machines, but also because those were the routine jobs that the first wave of automation was particularly good at automating. The second wave with robots, for example, was going a little bit more in the direction of more sophisticated jobs such as welding or painting and so on and so forth, which at first seemed harder, but that's also been achieved. So what happens then? Well, once those jobs disappear, you can either upgrade to really high, higher skill occupations, or you're going to be forced to downgrade or perhaps even not get employment at all. And given that we do not have an education system that provides sufficiently general skills, especially to people from low socioeconomic backgrounds, I think the cars were stuck against any type of skill upgrading for these workers, so them being pushed into orderly positions into low wage service jobs, I think, that was quite the expected consequence of automation in the middle of the job distribution. Right now, there's another phenomenon that's going on, and one of the things I find most interesting and stimulating about your work, also about the related reports of MIT's task force on the future of work or work of the future, is institutional change. There's a very interesting paper, which I'm sure you're aware of that came out of the Princeton Industrial Relations section, looking at the membership in unions from the mid-1930s, from the New Deal era, through its peak in the 1960s, early 70s, and decline thereafter, and showing a very tight inverse correlation, negative correlation, between union membership and inequality. I find it very interesting that a number of interventions, if you like, regulatory, statutory, institutional changes seem to be coming on the table. One of them, of course, is minimum wage loss. Another is the notion of collective bargaining, which necessarily would evolve in a different way with increasing numbers of gig workers, with individuals working for multiple employers, rather than lots of people working for one employer. But do you see, how do you see the opportunity on the need for these kinds of institutional changes around the labor markets as technology impacts them? I think it's critical that we think of technology as embedded in institutional frameworks, and you picked two of the more important ones, unions and minimum wages, but I think safety rules are very important as well. But I would also single out the institutional framework surrounding competition and innovation policy, what type of technologies we support, how much leeway we give to successful companies to muscle they worry around. I think those are also critical, both for technology and for labor demand. When we come to unions and minimum wages, they work, of course, in related ways, but they're very different also. I think I am completely supportive of a meaningful increase in the minimum wage. There are many reasons why wages in the bottom of the distribution in the U.S. are too low, and higher minimum wages would force employers not only to pay more decent wages, but also make other adjustments to use workers more effectively. Unions are more complicated, and there are several reasons for that. First of all, you know, the decline in unionization in the United States is both a cause and consequence of inequality. Unions were a largely manufacturing phenomenon, and other sectors were organized by manufacturing unions, and the manufacturing sector has been in decline, and especially with the automation technologies, the production workers in the manufacturing sector have been in decline, which are those are the ones that are losing their job. So there is an aspect to which we have to pay attention, which is that the decline of unions itself is a consequence of these things. But no doubt it does change in equality. But how we go about protecting workers through their organizations is also a very complex phenomenon. And let me sort of mention a couple of important issues. The first one is that you know, when they are at their best, unions can be even more effective than minimum wages. For example, work by my colleague Simon Jaeger and Benjamin Scheufer from Berkeley shows that work councils that have representation on boards and companies do not discourage automation. In fact, they may encourage automation in German companies, but when automation happens, they find ways of making the gains more broadly shared so that you don't have as many job losses and you don't have wage cuts in companies undergoing automation. So that shows the good side of unions. But I think the problem in the US is that, you know, compared to Germany or Scandinavia, employer union relations have always been more contentious, more conflictual. And when you have the more conflictual unions, for example, for Longshoremen, then that could be a way of capturing rents and blocking technological change even when that's efficient. And then the other challenge is, you know, because we're not going to have manufacturing workers in abundance in the future, how do we build worker organizations? I completely agree that building worker voice is absolutely critical. But how do you do that? And again, there are pitfalls to pick another bad case is police unions. I think I have very little positive things to say about police unions. That's not the model we should be aspiring to. So that really highlights to me that unions are not a uniform single phenomenon. We have to think of them in a more nuanced way. Well, I think you're absolutely right. And we're seeing some interesting, you know, emergent examples. I was struck in the the MIT report on work of the future of the discussion of the Unite Here union of service workers, particularly in Las Vegas, negotiating successfully to have a voice as in when technology is deployed, and also to collaborate on training programs for those who are displaced. The brotherhood of electrical workers also does that. But there may be clearly a need for experimentation. Absolutely. That's a very, very important point. Absolutely. Right. And led legislative framework, which given given, you know, the federal structure in the United States, much of this happens at the state level, it always has. And we may look for imaginative experiments to respond to this kind of challenge. I think that's absolutely correct. And just going back to the same point, and it's a new beginning. And we have to make sure that we can build in a way that is more conducive to cooperative relations between employers and employees. And it's a two way street. I think part of the reason why relations in the US have been so conflictually, because the US has a history of employers suppressing unions, suppressing voice. And so we have to avoid those pitfalls. Yeah, I was struck a couple of years ago. I remember that Volkswagen actually tried to bring in to invite union organization of its Alabama plant in the context of the right to work laws of Alabama and ran into a buzz saw when here an employer was trying to import, if you like, the German pattern of labor employer relationships and was blocked by statute on the books. But the again, as I say, the MIT report, I think, is very good reading on reviewing a range of opportunities that exist here. But let's shift gears a bit because you've now been speaking about a subject that's at least as contentious academically and beyond the academy as the issue of labor market institutions such as unions. And that is the question of, quote, redirecting technical change. The role for the state in guiding, subsidizing, supporting, shaping the manner in which general purpose technologies like computerization, digitalization are actually deployed in the markets of the economy. Would you address that set of issues? Thanks for raising that issue. And I think that's critical. And you are absolutely right. It is a controversial topic for too many economists. That is a no go area. Who are we to tell the creative entrepreneurs and innovators what to do? That would be a very natural fallback position for many economists. But let me make the case in two pieces. The first piece is that if there is any area of economic activity where externalities, spillovers, impacts on others are absolutely ubiquitous, it's innovation. There is no natural, neutral innovation. There is no natural path of technology. It's what technology is what we make it. Technology is what we imagine it. And every different piece of technology creates winners and losers. It creates opportunities for others and shuts down opportunities for yet other firms and other workers. So we really have to take those externalities and that richness into account, which means that we should give up the pretension that the pretence that somehow the market is going to naturally find a right path of technology. And the second observation is exactly where we started from, from a past of broadly balanced displacement and reinstatement, which was bolstering shared prosperity. We have gone to an era over the last 30, 35 years where we have much, much more automation and much, much less reinstatement. So there's a change in the direction of technological change. And this has coincided with much greater power of tech companies, both economically and ideologically. It has coincided with much less leadership from the government in terms of encouraging new technologies out of the box, thinking blue sky areas that, for example, were at the forefront of NIH and Department of Defense and other types of government support during the World War II and the post-World War II era. And it has also coincided with an epoch of much greater global competition, which naturally makes firms focus on reducing labor costs. So I think these three changes together suggest that there was a major transformation in the direction of technological change. And the pieces of evidence that we've talked about and many others suggest that that has been at the expense of labor, and especially at the expense of low and middle education labor, low and middle education men, especially. So I think we have to take those into account. And if, as some people in Silicon Valley and other positions of power would say, this is what technology wants, this is the path of technology that we have to follow. There's no other alternative. Perhaps then we would have to say, fine, it's creating this huge disruption. We'll have to find a fun way of dealing with it. We have to shut up and put up with it. But that's not the case. Technology is extremely malleable. We decide where it goes. So that's why I think this conversation about redirecting technological change is vital. And we are very late in realizing that. One of the distortions that you've discussed in your publications has been the fact that over the last generation, the tax system has evolved in effect to favor corporate expense, business expense on capital versus labor. You might elaborate on that a little bit. Thank you. That's the one thing I left off my list. I was hoping you would bring it out, but I didn't want to make it much longer. But yes, it will take a little bit of time to explain it, but I think it's worthwhile. So obviously firms look at their bottom line when they make their technological choices, both in terms of adoption and creation. But what matters for firms' bottom line is costs after taxes. So if you look at the US tax system, and here it's not a complete exception. There is a lot of parallel between what the US does in many other countries do. We treat capital more favorably than labor. It's always been so. And there are reasons that some economists would think are sacrosanct that you should always be more favorable to capital. I disagree with that in general. But as far as the facts go, capital, especially equipment and software capital were taxed at about 15% effectively in the 1990s. Labor, because it pays both the payroll tax and the income tax, was taxed around 25% at the margin. But since 2000, labor's taxation has remained roughly constant. Capital's taxation has fallen tremendously. That is for several reasons. First of all, capital owners tend to be richer and the higher marginal income tax rates have been reduced. There has been a lot of tax avoidance. Many companies declaring themselves to be S corporations and avoiding corporate income taxes. But most importantly, quantitatively, we have introduced these amazingly generous depreciation allowances, enabling firms to ride off their capital investments. So as a result, capital in software and equipment, the ones that are most involved in automation, are now taxed at 5% or less. Huge gap between labor taxation and capital taxation. What that means is that even if labor is better at performing some tasks, firms are going to find it profitable to automate because of the tax asymmetry between capital and labor. And that's been another powerful motive for greater automation. And it's exactly the wrong kind of automation because why? It encourages you to automate things where capital is actually not doing very well. You don't generate any productivity gains. So we are going back to these productivity effects. Right. Now, you've spoken explicitly about human-friendly tech. And how would you characterize that? And how might interventions sponsor, accelerate the development and deployment of that kind of technology? Well, I think that is the $1 million question. And to tell you the honest truth is I don't know for sure. You know, when you think of it abstractly, it is clear what human-friendly technology is. And historically, we can see and identify it. But moving into the future, it's a bit harder to see because technology, especially new things, new tasks, are hard to imagine. But human-friendly technologies, especially those technologies that create new tasks, new functions for labor. If you look at around you, the friends that you know, most people that I know today perform tasks that did not exist 60 years ago, even academics. You know, the job title is the same. But if you look at what we do using Zoom, using slides, searching information on the web, using complex computer programs, trying to reach a much broader set of people using technologies in completely different ways, combining tasks that are both at the research and public face of academia, all of those have evolved in very different ways. And if you look at other things like in the tech or in engineering or in design, they're completely new tasks. So if you didn't have these new tasks, labor demand for many, many, many of us would have been much, much lower. So those are one aspect of human-friendly technologies. But sitting, you know, in a similar conversation in the 1920s, we would not have been able to foresee exactly what these were. And in fact, probably one of the greatest economists ever came, you know, made exactly that error. And it's very rare because he was extremely clearheaded when it came to things like this. And in writing in 1929, he foresaw that A, these machines that were being rolled out would continue and automate a lot of jobs. And he also foresaw that that's going to improve productivity pretty steadily. He actually threw the number of 2% a year. Those were very much on the money. But then he couldn't envisage the counterbalancing new tasks that would keep labor demand healthy. And so he said, oh, we should all work part-time and we should all spend more time appreciating art because machines are going to do our jobs. So you cannot fault him because how could he foresee radiologists, management consultants, app developers, and, you know, Zoom technology, it's impossible. So that's why you have to allow for a diversity of voices. You know, when I was referring to the dominance of a few companies on a few vision, the biggest cost of that is because we preclude diversity and that diversity, people going in different direction, and the institutional framework and the support framework for those inquiries into new technologies being there is critical because we don't know exactly where specifically the most powerful human-friendly technologies are going to come from. You know, it's clear that we've got some very useful examples. The one that kind of comes to my mind relates to that subsector, that small and declining sector of the economy known as manufacturing. But the use of, in effect, augmented reality headsets for people who are putting wiring into aircraft, well, that kind of technology could be used for helping people who are employed to wire houses, to put plumbing in. In other words, it could be deployed at a level that we consider to be relatively low or moderately skilled jobs across a very large sector of the economy. And that leads me to think about how do we get these kinds of existence proofs and what role might there be for government. Again, a little bit in the model of what happened after World War II when the U.S. Defense Department took the lead in sponsoring all of the technologies that combined to create the digitalization of the world. So I think about maybe procurement is underestimated as a force that a state procurement, which is not driven by a narrow calculation of costs and benefits, but has a more strategic view, could be a way to demonstrate the viability, the usefulness of technology that's like virtual reality, augmented reality, is kind of on the frontier. It's what has been tried out in computer games. It gets some traction, but not necessarily, as they say, as we used to say in real life. So I think about there, perhaps there could again be a role for the state as a first customer, a collaborative and supportive customer for technologies that may not yet be ready for commercial prime time, but with sufficient Yeah, sorry, sorry, Bill. Yeah, I didn't mean to cut you off. No, I'm just going to say with sufficient practice in real applications, even if they're not immediately cost benefit positive, could pull them down towards being commercially viable? Yeah, I'm glad you raised this issue, both because the general issue is very important. And I'm going to put it slightly differently than you put it. And also because augmented reality, because I think that's actually one of the technologies that has great promise for being human friendly. But actually, we have already the proof of concept. There are many academic studies in engineering that show augmented reality is very powerful, and there are many companies, very small companies in general in manufacturing, that are using augmented reality quite productively. But I think its use is not universal or is not widespread enough. So I think government policy has a very important role. And I'm not going to close down any avenues of government intervention, procurement and others being included. But the way I would like to put it is slightly differently. Not that government procurement is a method of generating demand for some technologies. I think that's part of it. But the bigger picture conceptualization of that would be government leadership. So the government needs to show some leadership in encouraging certain areas. And that starts by government employing some of the best scientists so that they can actually spearhead that vision. Today, you wouldn't get people like John Von Neumann working for the government because the world has changed. And when you don't have more people like John Von Neumann working for the government, then it's much harder for the government to do that leadership. And then it comes with support for research, the R&D support, and then procurement at the tails of those two. So I think that ensemble is very important. And if we go just procurement, I don't think that would be productive or sufficient. And I think the augmented reality is a great example of things that we can do. But let me give you other examples, which could also illustrate what the problems are. Education. You can imagine AI being used in education in a very, very routine way. And that's the path that we're going to. We're going to use AI more and more to replace teachers. They'll do the test grading. They'll do some of the lectures and so on. But there's a very completely different way of using AI. You use AI to monitor which subjects and in what way different subsets of students are having trouble. And then you do much more targeted teaching and intervention. But you see the big difference between the two. One of them replaces the teachers. The other one requires more skill than more teachers. And which one are we going to follow? Well, I can bet you we're going to go the first path because A, the government is not giving schools money to hire teachers. The teachers are not getting much skill. How many teachers who are teaching in middle school are really at the forefront of the digital knowledge to be able to use AI technologies? And the business models of most AI companies is going to be much more in line with using AI technology for automating teacher work. So those are the places where we need this fresh thinking diversity leadership. There's a long story about the slowness of the diffusion of innovation into the economy. And what you just said brings back to mind something I experienced more than 40 years ago. I was in the early days of my transition from academic economics at the University of Cambridge into venture capital, what I call my 35 year sabbatical before I returned to Cambridge 15 years ago. But I was educated in technology and computing at a place called Xerox Park, the Palo Alto Research Center of Xerox, where the future was invented. And they had a program. They had a program running on one of their special purpose incredibly expensive, much too expensive for general use workstations called Buggy. Buggy was a diagnostic program for kids doing long division to work out why they were getting the wrong answer to make it easier and more efficient for the teacher to intervene and tell them about carrying or whatever might be the problem. That was 40 years ago. And that's exactly the kind of human friendly human augmenting technology that we need a great deal more of today. I find that a very remarkable example. Well, since you brought up the Cambridge academic environment of the 1970s and the 1980s, let me make another point here. You know, our conversation and much of my work, academic work has focused on the viewpoint of the American, British, French worker, the developed economy workers. And the argument we just went through is that the situation could be dire for them in terms of inequality, in terms of absence of good jobs, and ultimately, you know, several decades hence absence of many jobs. But the situation is even worse for the developing country where the majority of humanity lives. In the developing world, the competitive advantage is still labor. And if AI and digital technologies go more and more in the direction of automation, this would be an ideal case of something that Cambridge was at the forefront of developing for five decades ago, inappropriate technologies. So there is a sense in which the automation is a very much an inappropriate technology for the world. And nobody is taking that into account because none of those voices are at the table. That's a very good point. Our mutual friend, Danny Roderick, has made the point of premature deindustrialization in the developing world where the manufacturing was for 150, 200 years, the ladder through which peasant economies with very low productivity, even negative productivity, as my mentor Sir Arthur Lewis pointed out, climbed towards the high productivity, higher wage, more valuable occupations in manufacturing. But if manufacturing's automated, that ladder gets kicked away for the, let's say, post China developing economies. Absolutely. And I think that's something I am doing some work on. And there is some other interesting work, for example, by Marius Faber, who shows that robot adoption in the U.S. destroys jobs in Mexico because Mexican auto parts suppliers, for example, are now unnecessary. Many more of them are being automated in-house. Now, as we approach the end of our direct conversation, move towards Q&A, there's one other subject I just wanted to raise with you. We've talked a little bit about ourselves offline about the work of Frank Pasquale, who's got a new book out called The New Laws of Robotics, and where he is concerned particularly about, again, displacement versus augmentation of technology. But in his case, he's particularly talking about professionals. And one question he raises, and one question that I think we've observed a bit of, and I'm not sure how to think about it, but it has been the professionalization of more occupations through requirements for, for example, state licensing of people performing particularly service jobs, hairdressers, in a way that historically was never the case. And whether professionalization with the potential for collaborative, collective establishment, after all, it was only about 110 years ago that doctors were able to establish professional standards and get control over their environment in the context of state regulation. Whether through increased professionalization, the cost is perhaps rent capture by those who control access to the credentials, but the alternative benefit might be some way of shaping how technology gets deployed in those new professionalized occupations. Do you have any thoughts about that? Yeah, I think that's a great question, and I don't know the answer. There is some research among economists on licensing, and economists generally tend to be critical because of the rent extraction. My instinct is the same as yours that there are some occupations, dentists and medical doctors, and some other specialized occupations where you need expertise, that licensing may be very useful, and others in which it's just going to be a method of rent capture, and certainly the rent capture has proliferated. I think, let me decouple two things. One is ethical use of technology versus expertise. They're not the same thing, and I don't think they will travel together unless we create the right institutional framework. One example of extremely high expertise occupation today is AI scientists, big data scientists, people specializing in machine learning. Nobody would doubt the amazing skills of tens of thousands, hundreds of thousands of people now being trained and who have trained and are making huge contributions to the future of digital technologies. But we do not teach them ethics. We do not give them any sense of what are the social implications of the technologies that they are producing. So none of them or very few of them would bat an eye in putting all of their effort in creating new technologies that will create joblessness or monitoring or surveillance technologies that would empower companies or repressive governments. So I think it is partly our job and it's partly society's job, but partly also the AI profession to be more cognizant. So I think that's very important, but I'm not sure what your licensing would solve that. But let me then make another point related to what you've raised, Frank Pasquale's work and Shoshana Zabubov's work. I think one hope that we could have is that when things go really bad, the path of technology, the path of institutional change makes many people not keep up with the living standards that they are expecting, that inequality increases, dominance, power increases. Democratic governance will come to the rescue and there will be a backlash. But that's the problem that AI also makes surveillance, monitoring, misinformation, repression much more powerful. So it is weakening democratic governance as well. So we may be in the middle of a double whammy that we have chosen a path for AI and for automation digital technologies that are not creating shared prosperity and are weakening the institutional foundations that would self-correct such mistakes. I regret to say that that's one of the themes of the second edition of my book. The great reversal of the digital revolution that was sponsored by the state now striking back in effect and undermining the authority and capabilities of the state. But this is a good point at which to turn to some of the questions that have been accumulating. And the first one relates directly to something you were just talking about. And that's the role of schools and that I would extend to include community colleges. Is there an opportunity? Do we have existing institutions that with appropriate leadership and appropriate resources can play a role in equipping mere human beings to do a better job of coping with and taking advantage of the technologies that develop rather than being their victims? Well, that's a difficult question because I think at the high level of do we have the institutions? Yes, we have schools. We have public schools. We have community colleges. We have state universities. Those are the skeletal institutional structures that would be very useful. But the problem is that they're not doing the right thing. For example, there is growing evidence that in a digital era, social skills combined with technical expertise are very important. But if you look at our public schools, especially in the low socioeconomic areas, they are not doing a good job of teaching the social skills related to flexibility, to adaptability teamwork, and such things. And they're not necessarily doing a great job in terms of imparting technical knowledge either. The same is probably true. I haven't studied it in detail for community colleges. I think state universities, some of them are doing much better, but we need much more of it. So I think it's not that we do need to imagine new institutions, for example, in the area of worker voices I've talked about, we may need to imagine new institutions in the area of democratic voice. But in the case of schools, community colleges, universities, I think we just need to do a fundamental rethink of where they should go. Right. So moving into some of the further questions, here's one. What do you think of placing a specific tax on automation to encourage adoption, to fund retraining the workforce, rather than on replacement while taking into account the distributional consequences of technology adoption? I think that's a great question. I think we should have all kinds of instruments on the table. So I am not categorically opposed to an automation tax. But I would say two things. One is that it would be very hard to implement in the right way. We could tax robots, but robots are just one part of automation, a small part. How do you tax algorithmic automation? It's very difficult. But the second thing is, before we go to automation tax, I think we should correct our distorted overall tax system. So if you say, we're not going to touch the depreciation allowances, and we're not going to touch all of these S corporations that for just tax avoidance reasons are not paying any corporate income taxes. We're not going to touch all of the profits that many tech companies make abroad and don't pay taxes on. Then perhaps we will be forced to look at things like automation taxes. But correcting those is both more straightforward and much less controversial if we can just get around to doing it. Once we do that, and it's not enough, sure, let's come up with even more creative ideas, and I'm all for it. But I think there are these obvious things to be done. If we raise enough taxes on capital, then we may even reduce payroll taxes, and that would make even better when it comes to low-skill job creation. So I think there are many issues that we have to tackle first before we get there. You know, Devon, I actually what you just said rings a real bell with me because I thought for years that in response to climate change, if indeed we do move towards a carbon tax, the distributional impact of a carbon tax could be at least offset, if not more than offset, by using those funds to eliminate or reduce the tax on labor. Let's tax what we don't want, which is carbon, and stop taxing what we do want, which are jobs. Seems to be a fairly easy trade. But the political economy is complicated. You know, I wasn't there when those discussions were made, but people say, oh yeah, we couldn't introduce carbon tax, we had to go for cap and trade. So I mean, the big difference between cap and trade and carbon tax is exactly that with the cap and trade, you're not taxing it, you're creating the right price for carbon, but not taking the revenue. So there are political economy issues and it's complex, but yes, in principle, absolutely, we should be taxing carbon and we can use that revenue for other things. And of course, the payroll tax was introduced precisely because of another political economy issue. Roosevelt wanted to be damn sure that nobody could take away social security. And if you had to pay in to your own social security account, you had a vested interest, voters had a vested interest, as we've seen over the last now, almost 90 years, of resolutely defending social security and then Medicare funded by the payroll tax, distorting as it is of other values. Now, you spoke about the tax issue involving digital companies that can move their reportable revenues, reportable income to the lowest tax area. And we have a good, we have a question from Robert Owen at the University of Nod. In the light of both globalization and international externalities associated with technological innovation, how compelling and realistic are arguments for international standards and policy coordination given the complexity of the international distributive effects? And I think that relates to developing countries as well as the developed world. Yeah, I mean, I think they are absolutely critical. I think many of the problems that we're talking about, both directly and indirectly today, AI's direction, digital technologies, climate change, all of those are global problems. And global problems, of course, need global solutions and global institutions to back them up and international policy coordination is at the center of it. But it is very difficult to achieve as well. You know, for instance, when after the 2007-2008 financial crisis, when regulation of finance came on the scene, the most powerful argument that came from the financial industry was, well, if you tax us too much, we'll go to Singapore, we'll go to Shanghai, we'll go to Hong Kong. And that is always a threat unless you have international coordination. And of course, the reason why many companies in the US can avoid taxes on especially their profits is because there are lower tax jurisdictions such as Ireland, such as, you know, Cayman Islands and other places that, including the UK to some degree, Jersey, but, you know, with better international coordination, some of those problems can be ameliorated. Yeah, the French have been taking the lead with some broader European support now for some kind of tax at the revenue level, since revenues are much harder to move from country to country than reportable income. We have another question that relates quite directly to where our discussion focused. And this is from Yalena Reiljik Sapienza. The recent literature that investigated the nexus between innovation and non-standard jobs suggests that weaker jobs, low-paid, low-security, mean less innovation and productivity. You touched upon the role of unions in minimum wage, but you didn't mention the more general liberalization of labor markets from the 80s onwards that were intended to increase labor flexibility by weakening employment protection. Do we need a reversal of that movement? I don't know for sure. That's an excellent question. I think employment protection has always been much weaker in the US than in Europe, and the evidence is not overwhelmingly convincing, but whatever evidence exists suggests that employment protection is not a very effective tool, that it creates a lot of distortions, and it doesn't necessarily create a sort of broad prosperity for labor. On the other hand, especially in Europe in places such as France and Italy, the way that these distortions of employment protection have been tackled, such as through temporary jobs, has brought its own problems because now it has created a dual labor market and low attachment workers and so on. So I think it's been a veritable mess in that respect, and when you come to the US, I think the question bears on alternative work arrangements such as for Uber drivers or Lyft drivers, and that is a complex issue because there are very different types of drivers. There are full-time drivers and right now they are being deprived from services and benefits that full-time employees and other sectors enjoy, so it is important to protect them. On the other hand, there are temporary drivers and much regulation might be harmful by reducing the flexibility of these drivers. So we have to find perhaps a system that is sufficiently nuanced, that it differentiates gig workers that are permanent employees versus more temporary gig workers, so I think there's a lot of thinking to be done there. One proposal that's been made, in fact, is made in the MIT report that I referred to earlier, is expanding the reach of unemployment insurance to cover less than full-time employees, and again that gets into the federal state discussion in the United States, but it does represent one path forward. Well, I think the problem with that, I think that's a great idea. I think it's great from the point of view of workers, but it creates an asymmetry for firms. So when some firms can offload their responsibilities to the unemployment insurance system, and there isn't the right type of experience rating, then that's going to create a cost advantage for those firms, and those are going to be the firms that take the low road of creating zero power contracts, low attachment jobs, low productivity jobs. So we have to think about the full equilibrium of that. Absolutely. Now here's a question of immediate relevance, and it comes from Jack Gao who is at INET. Some say COVID has accelerated tech trends by a decade. What do you both want to see changed or happen once economies reopen? Daron, you go first. Unfortunately, we don't know. The only evidence I know on this is a number of surveys, and in all of those surveys, the answers are very much in line with what you suggest. Firms say that they have accelerated automation, or they are planning to do so, but we don't know the extent of it. I think it is very challenging to know what to do because when we come out, slowly come out of the COVID crisis, the sectors that employ low wage and lower skilled workers are going to be the slower ones to adjust, the hospitality sector, for example. Of course, one can say that could be an opportunity. We can try to find even higher productivity jobs for these workers, but I'm not sure that given the current commitment from the government, from the institutional framework, from technology, we really can hope that that's a feasible plan. So I think there could be very adverse consequences on the most vulnerable parts of the population. One response I will give to that is, as we come out of this, at least we can feel some comfort from the fact that leading figures in the new administration and the next administration include some of the best labor economists in the world, including, of course, Janet Yellen at Treasury, CeCe Rouse at the Council of Economic Advisers. We know we're going to have very thoughtful people looking at these issues in a way that I think it's fair to say it was unreasonable to expect during the last four years. I agree with that. So, Daron, any last minute as we approach the hour? Any last thought for our audience of young scholars? I just want to reiterate what I said is, I think we have come to understand that technology is critical for our prosperity, but I think we haven't completely internalized how much in our control and in our imagination technology is, and different parts of it have widely different implications for productivity, for distribution, for who benefits. I think we have to take those issues more seriously. Very good. Thank you very much. I think I certainly gained from this discussion, and I hope our audience did as well. For our audience, I now do want to say the following. The discussion can continue with me on the Inet Young Scholar Initiative. There'll be an open forum discussion. Please click the Young Scholars Initiative on the link at the bottom of the screen to join that discussion. Join us on Tuesday, January 12th, on Easter time at noon, where we will discuss our redistribution policies enough, moderated by the terrific Rana Faroohar of the Financial Times with two terrific economists, Gordon Hansen and Laura Tyson. If you haven't already registered, you can find the link to the right or on the Inet website. Please follow Inet Live and Inet Economics to hear about new episodes and information as we develop our series on the future of work. See you next week.