JPPE: Hi, everyone. I'm speaking right now with Professor Pascual Restrepo, who is an Assistant Professor at Boston University, whose research focuses on the impact of technology on inequality, as well as labor markets and economic growth. Hi, Pascual. How are you?

 

Restrepo: Hi, Julian. How are you? Thanks for inviting me.

 

JPPE: Thanks for taking the time to speak with me. So, the first thing I wanted to talk about is how you got into automation and inequality, and I saw that in your earlier work you tended to focus on the illicit economy--and from what I could tell, within that, the illicit economy in Colombia. And I wanted to know how you shifted from that towards automation and inequality. 

 

Restrepo: Yeah, absolutely. So I'm from Colombia, you know, and when I was doing my undergrad, I started doing my undergrad in mathematics, but then I got interested in economics. And the key topics in the public discussion in Colombia was always about violence, always about state capacity, always about corruption and political economy. And at the center of all of that, you always have these big illegal markets that were reinforcing, if not causing many of these things. So as a Colombian, it's natural to be interested in that, because it's a topic that is very close to our hearts, our history, and our future. And that's why I started doing research on that topic. Then when I moved to the US to start my PhD, the focus shifted a little bit. I guess that doing a PhD is a great opportunity to start seeing the world from a more global perspective, perhaps, and so I became more interested in problems that were perhaps not as important for Colombia. I mean, Colombia, I wouldn't say right now that automation, it's a big concern there. There's a lot of inequality, but for very different reasons. And so I became interested in these topics that I saw were very relevant for the developed world. And so I think that it's kind of nice that you get to do research on topics that you hear people talking about, right? So like, you open a newspaper, you read The Economist, and they're always talking about, you know, automation, inequality, technology, and so on, but when I was in Colombia, it was the case with illegal markets, so I guess that I'm just trying to follow that trail. You know, like, interesting stuff.

 

JPPE: So the shift towards a more global topic, with far-reaching implications. So, do you remember, was there a specific thing that you read? That made you interested? Do you remember the moment when you thought that this is something you wanted to focus on?

 

Restrepo: Yeah, I don't think that it was anything specific. I just think that it was like an accumulation of stuff. And you know, when I was starting my PhD, there was all of the fuss about Artificial Intelligence. Everyone was fussing about what it was going to be. And I was, like, listening to podcasts by, I don't know, Sam Harris and other people, or the book by Yuval Noah Harari, and I started reading all of these things, and I was like, "Oh, this is super interesting, I wonder if you can analyze this from an economics perspective," right? 

 

JPPE: Right.

 

Restrepo: And then, like, Daron--who was my advisor--and Daron apparently was also super interested in this thing, so we started this collaboration. 

 

JPPE: Yeah, yeah. So, I- and I think that this is one of the things that is so interesting about the topic, too, is that it is something that gets discussed in these journalistic spheres like The Economist or more central, mainstream publications, but also in academia. And also in a slightly more sensationalistic way by futurists and so on, because the name 'Artificial Intelligence' is kind of, you know, some people think of Blade Runner, or cyborgs walking down the streets. So I think cutting through that is very interesting, which is why one thing I did want to ask is where you stand on the question of automation. And I have four or five studies that you wrote that I just wanted to briefly mention, in case some people aren't familiar with that. One is a piece that you co-authored called "Demographics in Automation," where--and correct me if any of these takeaways are wrong, but--you showed that robots tend to substitute for middle-aged workers. In the paper "Automation of New Tasks," you showed that there is a reinstatement effect of labor, these new digital, labor-saving technologies. In "Competing With Robots," you showed that the overall impact of robot adoption on an industry tends to reduce the employment in that same industry--the number of jobs, at least in the short run. And in "Unpacking the Skill Bias," which I believe is the most recent paper you showed that there is this powerful impact on inequality, and that there is a reduction in real wages. And that productivity increases might not even be that high relative to inequality, which I thought was a very interesting point. So to the extent that you can, where do you stand on automation and AI?

 

Restrepo: Perfect, so let me try to reframe this question a little bit. One of the questions that you sent me earlier on--and feel free to tell me if you're going to discuss it later on, or where this is a good moment to discuss it, because you asked what's wrong with public discourse about automation, right?

 

JPPE: That was my next one, so we can do that now.

 

Restrepo: Perfect. So I guess that what would be useful for me is to tell you how I see public discourse, and where I see myself into the public discourse, and why I think that some of the perspectives that Daron and I have brought are different from the main views that you would find out there, right? Because you see that you are asking me where do I stand on automation. And I think that that's part of what's the problem with this topic, is that people want to divide themselves into two groups. There's kind of a false dichotomy, so either you believe in the robot-apocalypse: you know, robots are going to come and are going to take over all of jobs or you are a firm believer that this has happened before, and that we have already seen this, and technology is great, and nothing is going to happen. And I think that the reality's in the middle. Undoubtedly, technology is the only thing that has allowed us to achieve our standards of living. But I also think that there's no denying that technology sometimes achieves that at the expense of some groups of society. And I think that that's where I stand. Technology is a great force. Technology allows us to live better lives. But some technologies--not every technology, because technologies are all so different, right? Like, I wouldn't say that automation is the same thing as inventing new products, right?

Those two things do different things. Some technologies, like automation in particular, have this peculiar feature that the way that they generate productivity--that the way that they generate additional capacity to produce--is by substituting very specific types of human labor.

And those workers that get replaced and that get substituted, from their perspective, technology is a bad thing. Socially, technology is a good thing, but there are going to be losers--net losers. And I think that my thing is just to put the spotlight on those losers, and try to identify them and quantify those losses. The idea that technology generates winners and losers--I mean, people have that around, but we tend to think that the gains are so large that this is okay, we just need to redistribute. I don't know, the losses sometimes might be much more problematic than what you can compensate for with the gains.

 

JPPE: Right, certainly. I think even in the way that people phrase it: 'short term losses,' 'disruption.' Some people might say that there's a slight underselling of what the human costs of that disruption are. But I wanted to ask about whether or not AI is fundamentally different in your view, because you said it's somewhere in the middle. And it sounded like you were saying it's not going to be "a temporary loss of some jobs, but it's fine because we'll create new ones" or "a complete joblessness apocalypse." It might be somewhere in the middle. And that seems to me very much in line with what past labor-saving technologies did during the Industrial Revolution. You know, the classic example where you had a spike in the automation- or, a spike in labor-saving devices that increased inequality in the short run--and we can talk about the extent to which that was tolerable--but in the long run, those jobs came back. So do you think that it is fundamentally different from how devices in the past performed, or is it in line with that history?

 

Restrepo: Perfect, so that's a great question. You know, we talk a lot about Artificial Intelligence, but we don't really know what it does. Right?

 

JPPE: [laughing] Yeah.

 

Restrepo: Right, so I can talk about technologies that I know what they do. So for instance, automation of manufacturing via industrial robotics, or automation of white-collar jobs via software, right? So let's just start there, and then we'll speculate a little bit more about AI. So I think that industrial robotics and automation of white-collar jobs via software--the economic forces are very similar to what we saw in the past, let's say, with the mechanization of agriculture. Or the mechanization of textile production in England during the Industrial Revolution, right? You have very artisanal techniques where it took a bunch of people, kids and women, to weave and knit a particular piece of clothing. And then you come out with these machines where you just need one person to operate the whole machinery, and essentially what you did is you replaced all that artisanal and labor-intensive technique that used to be the main technique before. Like with agriculture, you see the same thing, right? Before it was like people with rudimentary tools, such as a scythe, or whatever who would reap the land. And then you have a tractor, right? And the tractor, there's just one person; and that person, the only thing that he has to do is drive the tractor. Nowadays they don't even need to drive the tractor, because these tractors are becoming increasingly self-driving tractors. And so, you can see that, essentially, we are moving to technologies that are more, more, more capital-intensive and that rely less and less and less on human labor. And I think that the consequences, to some extent, are similar. So like, many people say, "Oh no, we have a lot of mechanization in history, and we still have a bunch of jobs, so that's not bad." But then you look at historical records; that argument is not exactly right. Because for instance, England around the Industrial Revolution, there were about 60 years where wages were essentially stagnant. So yes, this technology created more bounty, they created higher incomes per capita, but most of this income went to the hands of capital owners: people who owned the machinery, people who owned the land. And those gains didn't trickle down to wages until after 60 years of these developments. 60 years is a long time. During these years we had a lot of unrest. We had many, many social reforms to appease, some of the unrest that resulted from these. We had the Luddites, right? That was kind of a response to all of these developments. So, you know, it was not a rosy- it was not an easy transition. Now, of course, that puts us into a fantastic path where eventually, people acquire skills; we came out with new industries, products, and so want to employ a bunch of people. But that was kind of like a choice, right? I mean, maybe that wouldn't happen. And so I guess that that's the next question: what are we going to be able to come out with? The jobs, the ideas, the tech force, the tasks, or all of the people that we displaced from manufacturing and all of the people that we are displacing from services. I mean, what are we going to employ all of these people, right? And perhaps, what if the future is one where we're going to have fewer jobs?

 

JPPE: You really think that there's a chance that the future could look like, on aggregate, there are fewer jobs that the economy requires.

 

Restrepo: Yeah.

 

JPPE: So in essence, the whole principle of efficiency gains creating jobs in new industries, that that principle might break down.

 

Restrepo: I mean, I think that that principle has worked in the past, but the fact that it has worked in the past does not guarantee that it's going to keep working in the future. I think that at least theoretically, conceptually, I mean, it's possible. I'm not saying that it's going to happen, but it's possible that we go into a future where the economy uses less labor. So you know, like, only 20% of the population works, and that's enough to supply all of the labor that the economy needs. I mean, this is a great thing if all of us only work one day per week.  But the problem is that we might be going to a future where only 20% of us work the entire week. And that's very different because the implications for inequality are quite different, right?  So I guess that that's kind of the challenge. The challenge to me is not so much whether the level of employment that you need to produce something is going to go down--for sure that's going to go down. I mean, we're going to produce much more with fewer workers. The question is, who's going to provide those hours of human labor, and who are the people in a position to benefit from that demand for labor that's going to be out there?

 

JPPE: Mm. So, I'd love to talk more about the social and political implications of what you just said, but first I want to ask about a pandemic-related question.

 

Restrepo: Yep.

 

JPPE: So, The Wall Street Journal just ran a headline where they said that, essentially, meatpackers were, all of a sudden, beginning to automate more and more labor, and that that automated technology was not necessarily doing as good a job as the humans were, but the coronavirus had essentially hastened that shift. And I believe there are other instances of that in manufacturing as well. So is this something that you view as bringing automation much faster?

 

Restrepo: Totally. One of the papers that you mentioned earlier on was this paper on demographics. And in that paper, what we showed is that a lot of what you see in terms of industrial automation--that is, automation by industrial robots and machinery in car manufacturing plants, for example--is driven by the scarcity of work. This is technology that responds to incentive. So what are the most automated countries in the world? Japan, Germany, Italy, countries where the population is aging very rapidly--where young workers with the muscle to weld a car are scarce. And so it's that scarcity of labor, in some sense, what has fueled a lot of this automation. And you can think of the pandemic as doing something similar; it's generating a scarcity of labor. Because workers--on the one hand, they cannot go to work because of either concerns or lockdown measures--but also, there's also some safety measures that might make some automation technologies more safe than human workers, right? So there's an element of that. On the other hand, you know, this pandemic also gave me another reflection that many of us- or, many people thought that humans were already kind of obsolete, and humans were no longer needed, and this pandemic kind of made me revisit that view because it does suggest that humans are still extremely important for the country. 

 

JPPE: [laughing] Yeah.

 

Restrepo: Extremely important, right? You take out the human element, and the economy completely gets destroyed.

 

JPPE: And human contact too.

 

Restrepo: Yeah. Absolutely. So, you know, it's not only about production, but so much of our economy is about humans interacting with other humans, and human contact that- it also kind of gives me some pause. Yeah, maybe, sure, maybe we're going to automate a lot of jobs, but maybe there's- you know, the economy's increasingly becoming more intensive in the sort of activities that, by their nature, they're just not automatable. Or we don't want to automate them because the quality of the goods that we are consuming then, they're dependent on human interaction, right?

 

JPPE: Mm. And certainly I think there are instances- I mean, the skeptics that I've talked to about AI have definitely brought up instances; they'll pull up a treasure trove of articles where someone says, "We can now automate kindergarten teachers."

 

Restrepo: Exactly.

 

JPPE: And then we could do some polling on the number of moms and dads that would be happy with a robot teaching their kids- or, their kindergarteners.

 

Restrepo: Yeah. Absolutely.

 

JPPE: [laughing] But, so, I did want to ask more about the social and political side of this as well, and I don't mean to be flippant with this question, but I wanted to know why inequality is necessarily a bad thing.

 

Restrepo: Great. I think that that's a great question, and I think that the answer is that it's not necessarily a bad thing, at least in my mind. Let me tell you the feature about current inequality that I think is absolutely a bad thing. I think that the problem with current inequality is not- I mean, you can have inequality for two reasons. You can have inequality because everyone's incomes are growing, but some incomes are growing much more rapidly than others. Right? Maybe there are some political, philosophical, or ethical reasons to be opposed to that type of inequality. Okay. But there, I don't think there's a very strong argument that everyone should agree that that type of inequality is very bad. Because that type of inequality might reflect, "Oh, maybe the people whose incomes are growing more, it's because they started a business." Right? You know, like Bezos. But Bezos has- is like a billionaire. But I would say yeah, if you ask me, I think that--well, maybe I'm more stating my case here, but--yes, he deserves to be a millionaire because-and a billionaire, because the product that he created delivers a lot of value, right? So I think that there's a chunk of inequality that is good and that I'm not opposed, at least, in terms of first principle. But. I do think that the nature of contemporary inequality is very different from that story that I was telling. Contemporary inequality is not just about some people who came up with great ideas, and their incomes grow a lot, and all of our incomes kind of grow just if we happen to work hard, and so on. Contemporary inequality, if I had to describe it, has this feature that we see groups of society who are worse off than before.  So it's not just how incomes are growing. It is that some incomes are falling in real terms. And for me, that is something that is just... the biggest contemporaneous problem. That we have people who are worse off than their parents, than their previous generation. So if, in the U.S., you are a person who has high school or less than a high school degree, who comes from a poor background, actually your parents were better off than you in terms of labor market income. And that's kind of the scary aspect of inequality, that it's not even progress, but sometimes it means no progress at all for some society. That's the reason that makes me worry. And those groups, of course--I mean, that's going to have political implications, everything. But just from a humanity perspective, I think that as a society we can afford- I mean, we should aspire for trying to make everyone gain from technology, everyone gain from globalization, there shouldn't be anyone who is worse off than a person like him or her 30 years ago.

 

JPPE: Mm. Mm. Yeah, and so, I mean, there is the data from Thomas Piketty and Emmanuel Saez that tries to quantify how that inequality is today; and from what I have seen, it shows that it's roughly where it was at in the early 20th century. And so when you think about automation in conjunction with the general financialization of the economy, the rise of private debt, the extension of credit in the economy, do you feel like now is a particularly vulnerable moment for a shock to inequality through digitally-enabled automation?

 

Restrepo: Yeah, absolutely. I think that part of the problem is that there's no safety net in this country. I mean, the safety net is kind of like- I mean we are seeing that with COVID, right? There's people whose finances are in an incredibly vulnerable position. And so imagine that you're in your 30s, you're in 40s, you have some debts, you don't own a lot of assets, and then they come out with this software that can do exactly what you're doing at your work, right? You are screwed! There's nothing you can do.

 

JPPE: [laughing] Yeah.

 

Restrepo: And then we turn to economists and the economists said, ‘no, because people can re-train, and people can relocate, and we're going to take that person and make that a coder, right? Or a software engineer.’ That's not going to happen! That's kind of like a lost generation there. And I think that we should care more about that potentially lost generation.

 

JPPE: Mm. And in terms of reducing inequality, I actually- a month ago, I got to speak with historian Walter Scheidel on his thesis and The Great Leveler, that inequality only gets reduced or leveled by mass military mobilization, civil war, plague, or government collapse.

And what was so striking about reading his book and talking with him was, it seemed to--on some level--be in opposition to a lot of the language in economics that seems to frame inequality and shocks as almost cyclical in nature, where there's a short run and a long run. And it seemed like Professor Scheidel's point was essentially that actually, there's been these exogenous forces in the form of, essentially, catastrophe, that have reduced inequality. And I'm wondering what your thoughts are on that.

 

Restrepo: Yeah. So I think that for economists in general, I wouldn't say that inequality's just a cyclical phenomenon. I mean, if you read the labor economics literature, there's a lot of literature kind of emphasizing these trends and it's long-run trends. So like, over the long run, what we are seeing is that someone who has a college degree is earning much more than someone who does not have a college degree. And that just keeps expanding and expanding and expanding, and decelerated a bit and so on. But you know, there are these big trends. I wouldn't say that it's only about shocks. But I do think that economists emphasize much more the role of technology than politics. Right? I mean, there are some fields in Economics that emphasize much more the role of politics, and I think that in these shocks that you're mentioning, what needs to happen is a big shock, so that it triggers political reform, and so on. And that perhaps hasn't been so much included into the language of economists. The other thing that I would like to say is that, while I think that thesis is kind of interesting, you also see a lot of variation across countries. When you look at the data that are countries that technology is kind of universal- I mean, I was telling you about, you know, the Germans were the ones who invented industrial automation, same as Japan. And they don't have as much inequality as we do here in the U.S., right? That's also kind of a choice; I don't think that you need to have a war to reduce inequality. But you need to have more progressive taxation, and I mean when you think about inequality,  that's the first thing that you should do! I mean, it's very simple. I mean, simple, technically.

Politically, it might be difficult, but the solution is clear.

 

JPPE: Right. And so, what do you think about proposals like a universal basic income, or a progressive basic income, or any variation of that? Negative income tax, and so on. 

 

Restrepo: Yeah, I think that anything that looks like a negative income tax, I think that would be greatly beneficial. Anything that looks like an earned income tax credit, that is subsidizing work for low-wage workers would be very beneficial. Also because it helps convince firms not to automate those jobs, so part of those jobs are subsidized. Anything like a universal basic income, I mean, I think it's- again, I like the idea; I don't like the idea of a universal basic income essentially substituting the whole safety net. I think that still having a government that buys insurance and that supplies all our programs that are more targeted to whoever it's useful. But I think that the spirit of all of these things is the same. We need a better safety net, right? And so, if you implement it one way or the other; if politically this is more feasible than this...I don't know. I'm fine. I just think that we do need some more safety net.

 

JPPE: Right. So the last question I wanted to ask is: looking to the future, what are some open questions that you think would be really interesting to look at, either on inequality alone, or on automation and inequality and so on? And also, if not questions, some areas of research that you think, thus far, have been untapped.

 

Restrepo: Perfect. So let me start with the first one. I think that one of the things that I find more puzzling about current technologies is that if you look at the labor market, you would conclude that there is a lot of disruption. You would conclude that there's inequality; the prices of different skills are changing the nature of jobs is changing; the types of jobs that we are posting are changing the skills that the labor market is valuing are changing. From looking at that, you would think, "Oh, technology is advancing at this amazing pace," but when you look at productivity, there's not a lot of productivity growth. And you know, this lack of productivity growth has been used by many to argue that automation, or that technology, is not a concern. So like, I saw that you did an interview with Paul Krugman at some point.

 

JPPE: Yeah. He brought it up.

 

Restrepo: Yeah, and I think he made that point. And I think that that point is kind of misguided. Because I think that one of the interesting things about automation technologies is that--and you already mentioned that this is something that we emphasize a lot--is that you can have automation without big productivity gains. And let me just give you an example of why that is the case. Imagine that I'm a worker working in a supermarket checkout machine, right? And then someone comes and invents a self-service kiosk, or whatever, right? That innovation is going to substitute me; but in terms of cost, how much is it going to reduce the cost for the supermarket? Very little! Because the worker was already very cheap; the machine itself is very cheap, but not as much! Right? Installing all of the equipment, programming, all of that--at the end of the day, the productivity gains is, I don't know, 1%? Something even smaller than that, that's just a small part of the cost? So at the end of the day, you unroll this technology, you deploy this technology, you substitute all of the checkout clerks, right?

So that's a chunk of the population, and what are the productivity gains from doing that? Not that big. They're not that big; it's just like, you saved a little bit on costs. So automation has this thing that as long as you save a little bit on costs, you're going to adopt it. So you can have the adoption of a lot of automation technologies that have very small efficiency gains. And I think that that principle has escaped current discussions about automation. People equate productivity or technology with automation. Automation is just one particular type of technology--concerning one particular component of productivity--that has this feature that it can generate big distributional impacts just by having small productivity gains. But I think that the interesting research question there is to try to understand, why are we adopting automation technologies that have such, very low productivity gains? And that our candidate- one candidate is perhaps taxation. Maybe we're doing this because we're taxing labor a lot so that's inducing us to adopt more automation, even if that's not very profitable, from an engineering perspective. Or maybe it's a cultural thing. So I think that those are the big, important questions. What determines the direction of technology? Are we going to keep focusing on automation, automation, automation? Or are we going to have a more balanced element of technologies that, in the long run, it's going to end up being more beneficial for all of those people--that I already discussed--that have experienced net losses in their income in the last 30, 40 years?

 

JPPE: Professor Restrepo, thanks so much.

 

Restrepo: Thanks. Thanks for having me, Julian.