Jesus Fernandez-Villaverde on Demographic Trends, Recent Macroeconomic Developments, and AI’s Implications for Economic Growth

Jesus Fernandez-Villaverde is a professor of economics at the University of Pennsylvania and is the co-director of the Business, Economic, and Financial History Project at the Wharton School of Business. Jesus is also a returning guest to the podcast, and he rejoins Macro Musings to talk about recent macroeconomic developments, the demographic issues facing the world, and AI’s implications for economic growth. Specifically, David and Jesus also discuss whether we needed the fiscal and monetary stimulus of 2021, the European inflation story, South Korea as a case study for global demographic trends, how quantum computing will may impact macroeconomics in the future, and more.

Read the full episode transcript:

Note: While transcripts are lightly edited, they are not rigorously proofed for accuracy. If you notice an error, please reach out to [email protected].

David Beckworth: Jesus, welcome back to the show.

Jesus Fernandez-Villaverde: Yeah, thanks for having me again.

Beckworth: Well, it's great to have you on. And I mentioned that you are a co-director of this Business Economic and Financial History Project at the Wharton Business School. And we've had several of your colleagues on the show before. We've had Peter Conti-Brown, Christina Parajon Skinner. We've had some of the PhD students on there, as well. We had Chris Hughes recently, co-founder of Facebook and one of your PhD students, and then also Kaleb Nygaard, who's also there. So, we've had a lot of people from that program. And of course, we've had you on several time as well. But tell us about that endeavor, that project. What's going on with it, and what do you hope to accomplish with it?

Fernandez-Villaverde: Well, Peter and Mark and Christina and many other people were interested in having a conversation about how economics and finance interact with history. And we felt that having a group, an initiative, or some type of institution where we could do this in a little bit more systematic way, could be good fun. And I guess that most of the kudos over here should go to Peter [Conti-Brown], who was able to go to the dean at Wharton and convince her to let us organize this thing. And it's a lot of fun. It's a small group, but we meet once a month more or less, and we talk about a lot of these issues. And I think good things are coming out. I'm actually trying to write a book with Christina Skinner. Hopefully, it will not take 10 years as some of my other projects, but it's really a lot of fun. And what I really like a lot is that we are similar along some dimensions, but along other dimensions we are very different. For instance, Christina, she's a law professor. Peter is more of a history/legal/financial historian. I'm more of a macroeconomist who happens to know one thing or two about economic history. So, the fact that we are sufficiently different, [there are] gains [from] trade. And that's what I think academic collaboration should be about, when I can learn from my colleagues, and hopefully my colleagues can learn a little bit from me.

Beckworth: Well, I should mention to the listeners that if they haven't gone back and listened to the previous shows with Jesus, I recommend you do that. But one of the comments I made back then, Jesus, I don't know if you remember this, is that you are the one person that causes me to lose faith in the idea of comparative advantage, because you're good at so many things. You're great at macro modeling. You have papers on how to solve HANK models, which are huge beasts. You also have papers on how to do the best times series analysis, the best way to identify shocks in macro, so you have all of that. Then, you're also a historian. You've written a book and you've written lots of history papers as well on global economic history. And now, you're doing this project with them. And we're going to talk about some demographic papers as well.

Beckworth: You seem to excel at everything you are doing. That really undermines my faith in comparative advantage, Jesus. Of course, you told me last time, which is a great answer, "Well, just imagine how more productive I'd be in one of those areas if I spent all my time in them," which is scary, to be honest with you. So, we are delighted to have someone as talented as you on the show. And you're well known for a number of things. But you're a great macroeconomist, and I appreciate all of the time you have spent on this show. And I want to tap into your mind and get your evaluation of what we've been through these past few years, and in particular the pandemic, and then the high inflation coming out of it.

Beckworth: But let me start first with just the real economy. And then, we'll talk about inflation, which is probably the most interesting [and] controversial part of this. But during the pandemic, and right afterwards, the economy contracted dramatically. If you go on an annualized quarter basis, I believe [Q2] 2020 collapsed about 30%. Then, a year later it was up about 30%, 35%. So, if you look at a time series graph, it's this real sharp, but brief, drop, and then really high, and you see it come down. And a question I have for you is this: did we need the additional fiscal stimulus in 2021 and additional monetary support to get the robust recovery we had? If you look at real GDP for example, it's pretty much back on trend, or it's near potential real GDP, or it's near where the CBO forecast would've had the economy now had there been no pandemic. And it's pretty remarkable. We didn't have any level effect. We didn't drop down like we did after great financial crisis. So, my question to you is, would we have returned anyways? Or did this additional fiscal and monetary stimulus play an important role, even though it may have cost us higher inflation?

Did We Need Fiscal and Monetary Stimulus in 2021?

Fernandez-Villaverde: My prior is that we would have gone back anyway. I wrote a paper, not on this, but on a related topic, where we were trying to estimate GDP at a daily basis using things like electricity consumption and things like that. And of course, you see that GDP on Sunday is very different than GDP on Saturday and is very different from GDP Monday to Friday. So, you will never say, "Oh, we need a fiscal stimulus on Sunday, so on Monday we go back to trend." This is not a bad analogy to what happened with Covid. Covid comes, everyone stays at home, think about a very extended weekend. And once we go back, we haven't forgotten how to produce things. We were just not at the office for three months. So in that sense, my forecast at the time, and I gave a couple of interviews, was that the economy was going to rebound quite quickly. Now, having said that, it is true that there were important things we should do, like help some of the households and some of the firms that could have suffered the most or that did suffer the most during this lockdown, but my impression is that we overdid it. And I'm relatively proud of having said that in real time, that we were overdoing it, and that we will probably be in a better situation if we would have been a little bit more careful. Now, I want to be very clear. I'm not saying that we should not have done any monetary stimulus. I'm not saying that we should not have done fiscal [stimulus]. I'm saying that, if we did 10, had we done 7, we will probably be in a better place.

Beckworth:  So, Jesus, the way I break this down is, 2020 was more understandable, even if not perfect, because we shut the economy down forcibly and we gave support to households. And maybe it was a little excessive, maybe it wasn't distributed optimally. But in 2021, it seems excessive, at least in hindsight. And some people like yourself [and] Larry Summers saw that. One way to think about this, the CBO, they estimated about a 400 billion dollar hole in the economy, and we were adding 1.9 trillion, the American Rescue Plan, and that just seems like a recipe... Larry Summers, using old Keynesian, hydraulic Keynesian would say, "Hey, you're going to overfill that bathtub with water,” if the bathtub is the economy. What do you think? Do you think 2021 was the pivotal year we went too far? Or was before then?

Fernandez-Villaverde: Yes. I recall, because I was given some round table in May, 2021, and I remember that, in that round table, I was the person who was the most skeptical about the latest round of stimulus. And I forecasted that we were going to have a lot of inflation. And the reason is perhaps very similar to what Larry Summers was saying. It's as follows: so 2008 to 2014, we have this financial crisis, and then the Euro crisis. There is a lot of fiscal and monetary stimulus, but it's an economy where there is a lot of shortage of aggregate demand. People don't want to spend. Suddenly, you have the government stepping in, and you have a more robust recovery without a lot of inflation. But in 2021, what is happening is, everyone is coming back to the economy.

Fernandez-Villaverde: The economy's still not at full capacity because we are still trying to handle some of the consequences of Covid, and yet, at the same time, a lot of households have saved a lot. It was in my own case, but I think it was pretty common, I didn’t go on vacations in 2020, so I have a lot of cash. And suddenly my wife and I say, "Oh, let's go to restaurants. Let's order a good bottle of wine," a little bit of revenge dining. And suddenly, not only am I doing revenge dining, but the government is doing all of this fiscal stimulus and the monetary stimulus. What were you expecting was going to happen? In some sense, my view of 2021 is that anyone with a good freshman introduction to macro will have pretty much forecasted this. This is not a situation where you need a supercomputer with a super complex model to forecast, and I'm a little bit disappointed that a lot of macroeconomists in 2021 didn't really see this coming. I think they were really a little bit too obsessed with the lack of inflation of 2009, 2014, and didn't really understand the fundamental differences between the financial crisis and the Covid crisis.

Beckworth: You're speaking to the inflation that was caused by the demand side. And I think you would recognize some of that was also supply side early on when [the economy] shut down. But when we think about the demand side inflation, is there a theory that you prefer? A macro theory? I mentioned old Keynesian, hydraulic Keynesian, or textbook Keynesian. You got new Keynesian, the gap between the expected and neutral rate going forward. Fiscal theory of the price level says look at the expected primary surpluses discounted to the present. Monetarist or quantity theory… is there a model that you would turn to help think through this period?

Applying a Macroeconomic Model to Demand Side Inflation

Fernandez-Villaverde: Yeah. The model I use when I think about this myself, when I'm doing exercises, and as an economist, so I always think about these things, some type of a mix between an neo-Keynesian model and a fiscal theory of the price level. On one hand, high [inaudible] deficits that pushes the price level up plus a very standard neo-Keynesian framework where there is both a narrative supply shock, the fact that there is still less people willing to work, still some global chain constraints. And yes, that the interest rates were so low, and the monetary authority was stating that it was going to be so low for so long, that people really embarking on a lot of consumption. I don't know if anyone has written this, but some type of neo-Keynesian fiscal theory of the price level is the best way for me to think about this.

Beckworth: Yeah. Doesn't John Cochrane say, every macro model has at least the government budget constraint in it, and from that you could kind of tease out the fiscal theory of the price level. And if you make the connection, once you look at that government budget constraint, it's hard to see anything but the linkages between fiscal policy and monetary policy.

Fernandez-Villaverde: When I was still an undergrad, and I read, *Some Unpleasant Monetarist Arithmetic* by Tom Sargent and Neil Wallace. And that's a classic that I will recommend all the audience to listen [to], especially if you're a graduate student, because it's so simple, even an undergrad could read it. And it basically says, "Look, at the end of the day, there is a budget constraint of the government. And monetary and fiscal policy are inherently two sides of the same phenomenon." And it is very, very difficult to come up from reading Sargent and Wallace without believing in some form of the fiscal theory of the price level, not exactly. If we want to do it in the way that John Cochrane does, or in the way that Chris Sims does, we can discuss at the margin, but really, *Some Unpleasant Monetarist Arithmetic,* I think, is one of the deepest papers that [has been] written in economics.

Beckworth: Well, it took some time for me to make that connection, I'll admit. But it's hard to escape it, again, once you see it. And I wrote a paper, man, this is after the great financial crisis, arguing that the Fed needed to do permanent monetary base injections. I was arguing very much in the lines of Michael Woodford. He talked about how, in Japan, the original QE, 2001, 2006, it was almost as if the market expected the Bank of Japan to reverse it. And they did. They reduced the growth of the monetary base. So, a number of us, myself included, were saying, "We need permanent injections, ones that are expected to be permanent." And I wrote several op-eds. I wrote a journal article on this. But only later did I realize, how did you get a permanent monetary base increase? Well, you got to look through the government's budget constraint. That implies, in the future, the government is not going to increase taxes or run a deficit. It requires some offset or some corresponding action on the fiscal side. And it's important to think through, what actually is a permanent monetary base injection?

Fernandez-Villaverde: Yeah, when I was an undergrad, again, I mentioned… and I had a great professor, David [inaudible], and he always emphasized that 90% of economic policy in general is just about understanding two things, that the government budget constraint exists, and second, look at the variables always in logs, not in levels. He said, "If you get those two things right, 90% of your economic policy is going to be right and 90% of your advice is going to be right." And he was very involved in economic policy himself, and I think he was absolutely correct on that one. Just thinking about the government budget constraint gets you very far, really, really very far.

Beckworth: Okay, let's take that idea and our discussion of inflation and apply it to Europe. We're talking about the US. But you know a lot about the European economy as well. You're from there. You were just there at the conference with the ECB. What is your story of what's happened to inflation in Europe over the past few years?

The European Inflation Story

Fernandez-Villaverde: It's actually very similar to the US. Europe also engaged in very expansionary fiscal and monetary policy. And in addition to it, in Europe we had the big [inaudible] of energy. One of the most important geopolitical facts that most people don't interiorize these days is that the US is already a net exporter of energy. Since around 2021, early 2022, the US actually produces more energy than it consumes. Also, the US could literally just close from the rest of the world and have enough energy. At the margin, some of the type of oil that the US produces is not exactly the oil that the US economy needs, but that will be a minor adjustment. Europe, in comparison, is an enormous importer of energy, and you have the Ukraine war. You have Russia invading Ukraine. And suddenly, you have natural gas, which is an enormously important component of the energy mix in Europe, going through the roof. And that imposed a lot of constraints on inflation in Europe. And that really has dominated the dynamics.

Fernandez-Villaverde: And things turn out to be a little bit better than expected. Credit should be given, in particular, to Ben Moll, he's a professor of economics at the London School of Economics, and some of his co-authors, and I apologize now to them because I don't remember all of their names, but Ben and his co-authors very quickly pointed out that Europe actually had a lot of margins of adjustment away from Russian gas, and that this was going to be a few months. And after those few months, inflation was going to moderate. And that's exactly what we are seeing now, that Europe was able to move away from Russian natural gas much more easily than anyone expected. And basically, what we have had is one year… well the Ukrainian war, as of today I think, is we are in the day 501, or something like that. So at least during the first six, seven months, really imposed huge costs on the energy markets in Europe. And that was something that was quite different from the US.

Beckworth: Okay, so a lot of supply side disturbances, energy issues involving the inflation, more so than the US. So, what role is there for the ECB? Here in the US, we often talk about the Fed falling behind the curve in 2021, 2022. Has the ECB done an adequate job?

Fernandez-Villaverde: I think, they also fell behind the curve.

Beckworth: Okay.

Fernandez-Villaverde: And I have also written about that. I think that, unfortunately, central banks these days are suffering a little bit from mission creep. They are getting into more and more areas that are not their core goals, their core targets. And that means that the principle policymakers within the central banks tend to get distracted, and I think they are dropping the ball. And I think that the European Central Bank dropped the ball for at least six months, and that the interest rate path could have been better. On the other hand, it's also true that now inflation seems to be going down much faster than anyone expected, and things hopefully will not look that bad, although the Eurozone is already technically in a recession. So, it may be the case that, when we give the final grade, we can still give the European Central Bank a B, I will say. But I rather would have had the European Central Bank with an A than with a B or a B minus.

Beckworth: But if nothing else, lessons learned for the next big supply shock we face, moving forward. Let's transition now into your papers on demographics. You have three articles that you've written among others, three that I'm going to use for the show, and we'll provide links to them. I know they're on your webpage, but we'll provide links to the readers in the show notes. And the three articles are titled, *[The Demographic future of Humanity: The Trends]* The second one is, *The Demographic Future of Humanity: Economic Challenges.* And the third one is, *The Demographic Future of Humanity: Social Change.* Let's start with the first one, just outlining the facts, what's happening to demographics. The first one's called, *[The Demographic Future of Humanity: The Trends],* And in it, you mentioned that we're about 8 billion, now. We'll peak at about nine and a half billion in 2050 to 2060. But then we're going to fall after that. And you use as a case study, South Korea. Maybe walk us through South Korea, and why it's so useful to look at it.

South Korea as a Case Study for Global Demographic Trends

Fernandez-Villaverde: I find the case of South Korea interesting. Because I feel it's a little bit the canary in the mine, the future for a lot of us. It's perhaps, because of some social and cultural norms, a little bit of an extreme case. And I don't think the US or other countries are ever going to be there. But at the end of the day, South Korea is in a path to lose, I will say, 80% of its population over the next century, unless they have massive immigration, because the fertility rate now, in 2023, they are going to be around 0.7. That basically means that every woman on average is going to have 0.7 kids. A way to think about this is as follows: imagine that you're in your high school class. We were 40 kids in my high school class. Let's suppose there are 20 women and 20 boys, 20 girls and 20 boys. That means that, on average, that class of 40 is going to have 14 kids. Okay?

Beckworth: Wow.

Fernandez-Villaverde: You have 40, and you substitute with 14, and just keep doing this a couple of generations-

Beckworth: Right. You get really small.

Fernandez-Villaverde: ... and you get really, really small. And I heard people saying things like, "Oh, well, but you can fix that with immigration." And I always have two answers. Well one is, "Yes, but that will basically mean that, in two generations, South Korea will not be South Korea, it will be something different," which may be perfectly fine, but as a society you need to understand that. And second is that, I'm going to make a claim which is, 2023 may be the first year in the history of humanity where the fertility rate of humans fall below replacement rate, okay? And I don't think people have really interiorized that. Let me walk you through some numbers.

Beckworth: Yes.

Fernandez-Villaverde: A lot of listeners probably have heard that the replacement rate is 2.1. That's kind of the number people always quote. But that's actually not true. That's true for the United States. That's true for western European countries. It's actually not true for the planet. And the reason is because 2.1 is basically the following idea: a woman has 2.1 children, on average you have around 105 boys born for 100 girls. And not all the girls survive to their fertility age. So, you need a little bit more than two. The problem is, on the planet, there is still a lot of selective abortions, particularly in China and in India, but in many other Asian countries as well, which means that, on the planet, there are not 105 boys born for each 100 girls. There are actually around 108, 109. And also in Africa, where a lot of the current births are happening, infant mortality is still relatively high, which means a lot of those girls will never go to the fertility age.

Fernandez-Villaverde: That tells you that the replacement rate for humanity is probably around 2.2, maybe 2.25. And the fertility rate for humanity in 2023 will also be around 2.2. We are right at the corner. It may be the case that when the data finally comes in in 10, 20 years, it may have been 2022, maybe 2024, but according to my calculations it will be 2023. So for the very first time since 200,000 years ago, we are not reproducing ourselves. So yes, in the past population went down, because there were famines and hungers and wars and epidemics, but there were always a lot of births. Now, the planet… and again I want to emphasize this, this is not about the US. This is not about rich and advanced economies, this is about the planet as a whole. We are not having enough kids to replace ourselves. Just fast-forward 30, 40 years, and that implies that the population will start falling around 2060, 2065. Again, there is a lot of uncertainty. And South Korea is just 40 years ahead of the pack, and that's why South Korea is so interesting.

Beckworth: So many questions there that this raises, so I'm going to cite some numbers from your paper just to make this very concrete for our listeners. South Korea currently is 51.8 million, and you estimate it could hit as low as 12 million when it's all said and done. That's huge. That's mind-blowing. You just begin to think of all the empty buildings, right? If you're housing 52 million South Koreans, and now we only need 12, and we'll get to this later, but that has huge implications for costs, for maintenance of capital and infrastructure, things like that. You also alluded to China, and of course China's been in the news because... was it this year that China is beginning its first absolute decline in population or was it last-

Fernandez-Villaverde: Yeah, 2023.

Beckworth: Okay. So, they’re currently at 1.4 billion, and you expect them to probably settle down around 700 million. That's cutting it in half. That’s a huge… I just can't imagine all of these stories about, oh, they've poured more concrete in China the past 10 years than the US did the past 200 years. All of this capital structure, how is it going to be maintained when you cut the population in half, or more?

Fernandez-Villaverde: But I can tell you because I have already seen it. I'm from Spain originally, as probably most of your listeners can tell from my accent. Spain also has a very low fertility. And I'm from the north. And the north is the place where fertility fell first. And in addition to it, I'm from a town, where my family is from, a town that also suffered a lot of out migration. It was coal, basically. It had a lot of coal mines, and the coal mines shut down. So, if you go to the hometown of my father, now you can get an apartment, a two bedroom apartment, for around 10,000 euros. Now, it's not in great shape. It's not a fantastic apartment. But it's 10,000 euros.

Beckworth: Wow. That's cheap.

Fernandez-Villaverde: Yeah.

Beckworth: Very cheap.

Fernandez-Villaverde: I know. I know. I know, and if you want, I will be more than happy, there is a Zillow, but for Spain, which everyone uses, I'm happy to send you the link. You will see that even the best apartments in the town go for $150,000, $175,000. If you show up there with $250,000, you cannot buy anything. Not because you don't have the money, just because there's nothing as expensive. And what do you think is going to happen in most of the world when… Now, of course there is going to be, and I try to explain that, this is not going to be the case in a few super highly desirable areas, okay? The problem when population goes down is not that it goes down uniformly everywhere. So, in South Korea, Seoul is actually still gaining population. And I can see the fancy neighborhoods of Seoul with the great restaurants and the great bars, the price being very, very high. But I can also see a small rural area in South Korea in 2040 where you basically can buy a house for close to zero.

Beckworth: Yeah, you mentioned that in your third paper. We'll jump there and talk about that right now, *The Demographic Future of Humanity: Social Change,* that there'll be a big exodus, even more so from the rural parts of the world to the urban centers, particularly for young people. Because that would be a better place to live, right? There's things to do. And then you also note that living in the city reinforces the downward decline in fertility, because the opportunity cost is higher there. This just exacerbates, makes things more so. And of course, the other points you bring up are that, when you have this, we're probably also going to have an aging population, low fertility. So, more and more older people per younger for younger person. So, we've got to support more older people with fewer able-working bodies.

*The Demographic Future of Humanity: Social Change*

Fernandez-Villaverde: So, I'm working on a paper right now with Gustavo Ventura. It's called, *The Wealth of Adult Nations.* And we do the very simple following exercise: take the GDP of Japan, and instead of looking at GDP per capita, look at GDP divided by people between 16 and 65 years old, and normalize GDP per adult in Japan to one in 1995, and to the US also to one; you do the same exercise. Which country has grown the most, Japan or the US?

Beckworth: I'm going to guess they're similar, because I read your paper.

Fernandez-Villaverde: Basically, yeah, Japan has grown 1% more. But just think about it. You go to any conference, and in macro you are going to find 20 papers, "Oh, Japan. The monetary policy of Japan is wrong, or the fiscal policy of Japan is wrong, or this or that of Japan is wrong." My argument is, all of these papers are useless, sorry to say it in that way, are completely worthless. Basically, there are 10 million workers less in Japan now than in 1995, what do you think is going to happen? How can you grow if you have fewer people? Even if productivity still grows at the historic level, if every year you lose 1 million workers, you are never going to be able to really break above 1% or 1 and 1/2% a year. And what is the Bank of Japan going to do about it?

Fernandez-Villaverde: And the argument in the paper is that you can do this through a lot of different countries, and you really get a very different view about the macroeconomic performance of countries over the last 20, 25 years, which is if there are fewer Italians, there are fewer Spaniards in working age, I'm not talking about… A lot of them are over 65, but they are not working anymore. The way you want to think about macroeconomic variables, macroeconomic aggregates is very different than the way you would like to think about the world if you are looking at per capita. And our point is that, more and more GDP per capita is going to start being a very misleading indicator for things like judging economic growth. It will still be good to think about issues like welfare, but certainly the headline, "Wall Street reports, GDP per capita grows 1%," I have started to think that this is a less and less useful statistic to look at.

Beckworth: Okay. You touched on Japan, and I want to ask a question about this. And this actually jumps into your *Economic Consequences of Demographic Decline* paper, but Japan is the avant-garde of the canary in coal mine. You mentioned South Korea, but Japan's been on this path for a long time, right? And one of the features of Japan… you just mentioned GDP per capita has actually been respectable and even better in some cases than other advanced economies, but one of the other defining features of Japan has been low inflation and low interest rates. Now, I know many stories that one could tell about why that is happening, aging population, maybe they're more risk averse, they have safer assets in the portfolio on the margin, or maybe they live longer so they save more. So, there's this story you could tell. There's another story that one could tell, and that story is told that, as we age we'll spend more, that we'll start tapping into some of that savings. Also, there'll be less of a young global force to join the labor force, and put down wages around the world. And I'm just wondering where you would come down on this. Is Japan the place to look for the future of interest rates and inflation? Or is it not necessarily the best example for us?

Fernandez-Villaverde: I think I have refereed a lot of papers that try to make similar arguments. And my standard referee report is always, "Well, the problem, to tease that out, is that Japan is integrated in the global economy. So, the capital markets are integrated."

Beckworth: Okay.

Fernandez-Villaverde: Really, I think it's having an effect on the aggregate wall or the aggregate advanced economy real interest rate. Teasing that out just from Japanese data is tricky. But yes, you're absolutely right that there are two forces. "I want to save more because I'm going to live longer" versus "I'm old and I'm spending already what I have saved." And which of the two mechanisms matter the most depends a little bit exactly about when your fertility went down, how you're evolving your demographic structure. What people need to understand is that demographic choices have a very, very long run impact. Let me give you another example that maybe makes this point in a slightly different context. I have also been very involved in education reforms in Europe, trying to make education more suited to the modern world. And what I always try to explain to people is, if we change the educational system today, the first student that will graduate from graduate school with a new education system will only come to the labor force in 25 years. So, anything that we do in 2023 will not have any impact until 2050. Okay? So, in some sense we haven't seen anything yet.

Fernandez-Villaverde: And I hope it's clear in the three pieces you are mentioning that I try to say that I think about what I'm saying a little bit more as educated guesses or conjectures than tight predictions, because we have never been here before. We don't really know how societies are going to react to constant losses in population, to gigantic migration movements, to a variation of the population. Let me give you another example. We know that older people vote differently than younger people. As the population ages, this is also going to have first order political consequences. What are going to be those? Well we don't know. We can guess. We can make some conjectures, but this is going to have a very important fit. And in Europe you can already tell that parties these days do not really pay a lot of attention to the needs of young people because there are not that many of them. You don't win elections to Congress in European countries by making your 25-year-old person very happy, because there's not that many 25-year-old persons. But there's a lot of people [that are] 65, and political parties are very smart, and they respond very, very clearly to the fact that now the median voter is maybe in the mid-50s.

Beckworth: Yeah, big changes. And you note this in your paper. You talk also about the US, the electoral map has changed dramatically, and will continue to change going forward. You also mentioned some other challenges that have emerged from this development is that universities, hospitals may not have the support they need as populations dwindle. We already know this about universities, enrollments are going down, and we know they're going continue to go down. And there's going to be a huge shrinkage and consolidation of higher ed in the US.

Fernandez-Villaverde: No, totally. And what I try to explain to people also is that this interacts with the fact that universities have hierarchies. What do I mean by that? Look at my own university. I teach at University of Pennsylvania, relatively fancy ivy league school, yada yada yada. Now, the cohort of people wanting to get into the universities are smaller, that's not going to really affect Penn, because the only thing we are going to do is we are going to lower our admissions standard. So, instead of you needing, let's say out of 100, be of quality 90, we will only lower to 85. So, who are the 85 to 89? And I hope no one gets offended, but they are the kids who will currently go to Penn State. So, we are going to be poaching the best students that now go to Penn State.

Beckworth: Yes.

Fernandez-Villaverde: What is Penn State going to do? Well, they are going to turn around, and they are going to go to the next university in the ranking. The problem is, when you go down three or four levels, there is nothing left. Okay? It's not that all universities are going to have 20% or 25% less students. It's that the top universities are going to be perfectly fine. The bad universities are going to be in a really deep problem. Again, I'm working with a former student of mine who is South Korean. And I taught him this model, and he actually studied… his undergraduate was one of the top elite South Korean universities. And he told me, "Oh, no, no. This you can already tell, we have already had an exam." And he says, "The type of grades that undergrads get today to get into my university," think about this like the Harvard of South Korea-

Beckworth: Yeah.

Fernandez-Villaverde: ... "They will have never been accepted 20 years ago." And in addition to it, this also has another loop which says, since you know it's much easier to get into a top university, you put less effort. So suddenly he says, "The type of students and the type of effort you need to get now to get into Seoul National or Korean University is totally different than it was 25 years ago." And we are actually trying to get data on this to see if we can document this beyond the impression of my student who says, "You can tell, and people talk about this." And that's what is going to happen. Basically, you are going to get that the big campuses… let's take Ohio. The flagship campus of Ohio State is still going to have students, but some of the local campuses of Ohio State, they’re just not going to have students. No one would like to go there. And then, what do we do? Then, good luck going to your local constituency and saying, "We are going to close this local campus." People are going to hate that.

Beckworth: Yeah. Well, let's move to another challenge this demographic decline brings up, and that is differences in fertility rates among secular versus religious people. And I think we're all aware of, in Israel right now, a lot of the tension is grounded in the fact that the Orthodox Jews, more conservative Jews, their families are growing rapidly, and the secular Jews are like the rest of us, having few kids. And so, they can see the writing the wall. The future is clear, if things continue, the orthodox conservative Jews will have more influence, make differences in how Israel is run. Even in the US, with certain religious communities in the US, they have bigger families. Now my question to you is this, though: outside of Israel, will this be a big deal? Is it going to be like in the US or some other European country, where the religious communities will grow enough to change how policy is implemented? Or is it just an interesting fact that the religious communities will tend to be having larger families as the rest of us have smaller families?

Religiosity and Idea Generation: Fertility Rate Impacts

Fernandez-Villaverde: So let me give you an example of the US where this is already binding, Pennsylvania. The Amish have grown so much that, at this moment, if you want to run for governor or senator in Pennsylvania, you really want to sit down with the leaders of the Amish community and ask them, "What do you want? What can I do for you to today?"

Beckworth: Interesting.

Fernandez-Villaverde: Remember, there is also a second point about religiosity that people tend to forget. Religious people vote more. So, it's not only that they are growing in relative size, also their participation share is very high. Okay? And let me give you a second example. This is not as much as religion, because it also has a bit of an ethnicity aspect to it. Ireland was divided, the island of Ireland, between what became later the Republic of Ireland. First, it was a free state. It was 26 counties. And of course, the six counties in the north of Ulster. When that happened, Protestants were around 60% of the population and Catholics were a little bit less than 40%. Probably in the next census, Catholics are going to become a majority. So, over one century, Catholics had just a little bit higher fertility than Protestants. And the reason I mean here is ethnicity is because in Northern Ireland, when you say Catholic, it doesn't mean you go to mass on Sunday. It means that you are from the Catholic neighborhood.

Fernandez-Villaverde: And  that has dramatically changed the politics of Northern Ireland. And you see over there, you see the electoral districts, how the ones that used to be Protestant, more and more now are becoming Catholic, because basically Catholics vote for Catholic MPs and Protestants vote for Protestant MPs. And places where Protestants would win, now they're being won by Catholics. And that's really completely changing the future of Northern Ireland. In that sense, I think that people really underestimate these type of effects.

Beckworth: Wow. Okay. So, it is more than just an interesting observation. It has big political ramifications moving forward. Alright, one more thing on demographics, and then we'll move to AI, and that is the importance of economic growth. And in particular, I want to focus in on ideas generation. You've outlined this point, which to me has always been an important one. You often hear some commentators say, "You need to have one kid because of the climate." And I always say, "Well, look. If I have more kids, there's a greater chance one of them is going to be an Einstein and find a solution for climate change." The more people you have, the bigger this distribution of IQs… you got to have so many million people before you have an Einstein, right? And so, it's important to have people, ideas, brains. And the flip side of that… that's the supply side. The demand side story is, the bigger the market, the more specialization, the more productivity, the more innovation and productivity gains. So, there's both supply and demand side story to idea generation with population growth. And yet, what you're telling us is that this is going to be something that, when we look forward, we won't have that force there, that energy there in that area because we're going to have decline in population.

Fernandez-Villaverde: Yeah, exactly. Let me give you a couple of examples. There are several papers that already make the following very simple point: you actually look at the rate of entrepreneurship in the US, and it has been going down a lot. And again, a lot of people say there's maybe fiscal reasons, must be the regulation. So, Hugo Hopenhayn at [UCLA] has a very simple exercise. He basically shows that 28-year-olds are as entrepreneurial as 30 years ago. There are just fewer of them. Let's imagine that one of each 100 20-year-olds becomes an entrepreneur. If you only have 75%, as many as in 1980, you are just going to have 75% of the same number of new firms than in 1980. And again, what are you going to do about this, okay? And that's the same thing as you were saying.

Fernandez-Villaverde: It's a very silly example, [but] since I'm European, I need to mention some sports. Just think about soccer players. If your cohort is 3 million people, the chances that you'll have a fantastic soccer player in that particular year is very high. If your cohort is 1 million people, you have only one third of the possibility of having a top soccer player. And again, what consequences [is this] going to have for humanity? I think they are going to be quite fundamental. What I try to argue is that this is going to be particularly important, because I'm convinced by the research that argues that ideas are getting harder and harder to get, because everything that is in some sense obvious has already been invented.

Beckworth: Yeah.

Fernandez-Villaverde: Being very, very smart is more and more important than ever and ever, and I see this even in economics. I talk with my graduate students and some of them express frustration about how difficult it's to come up with an innovative paper, because every time they come to my office and say, "Oh, I'm going to write a paper about this." And I say, "Well, Smith and Williamson brought that paper in 1995." If you were in 1960, that was way easier, because people were not really writing those papers, nothing had been written, or much few fewer papers had been written. So, new and newer ideas are getting harder and harder to come up with. The fact that there is less and less of us and the young people trying to come up with new ideas is just going to slow down economic growth. In that sense, I'm relatively pessimistic.

Beckworth: So, ideas are getting harder to come by because we already picked the low-hanging fruit, number one. And number two, with fewer people, there's fewer idea generating. So, this is a compounding effect which is going to affect future productivity growth; a very sobering message there. Well, let's move on from that, in the time we have left, to artificial intelligence, AI. As you know, it's been a big thing. I've had a few guests on to talk about it. And you've written several papers. And we're going to get into a few in a minute. But I want just put a plug out there for one of your papers, which is sort of AI-ish, but it is really unique, and I'm not sure I really understand it, but I want to put a plug because sounds like macroeconomists are doing quantum physics. But here's the paper. It's an NBER working paper. It's titled, *Dynamic Programming on a Quantum Annealer: Solving the RBC Model.* And you have an interesting abstract. But are you doing quantum physics with macroeconomics now?

Quantum Physics and Macroeconomics

Fernandez-Villaverde: Yes. If you go to my office, you will see that I have three or four books with titles like Quantum Mechanics, Quantum Dynamics, and things like that. No, but, this is not as crazy as it sounds, okay? Let me give you a little bit of background. The computers that we all use… so we are recording this interview on a computer, and I guess you also have a computer, are what are known as classical computers, okay? And those are based on classical physics, basically zeros and ones. We have a current in the processor. There is an alternative way to design computers that is using quantum mechanics, and those are called quantum computers, okay? And those take advantage of some of the properties of quantum mechanics and quantum physics, which are really fascinating and allow us to do things that are completely unheard of.

Fernandez-Villaverde: Now, quantum computers for the longest time were just experimental things. IBM had one that could do very, very simple things. There is a version of quantum hardware called quantum annealers, which are kind of a baby quantum computer that can do only a few things. And the good thing about these quantum annealers is that my co-author, Isaiah Hull, and I could use one of them. So, what we did was to say, "Well, for the fun of it, can we compute a standard dynamic programming problem in economics using one of those quantum annealers?" And that has two challenges. One is, you need to think about an algorithm that takes advantage of this type of quantum phenomena. For instance, there is something called quantum tunneling, which is quite fascinating, but I'm not going to bore the audience about some of this, but for those interested, check on [Wikipedia], quantum tunneling. It's absolutely fascinating. So, we devise an algorithm that takes advantage of the quantum mechanic aspects and then we actually ran it on a real quantum annealer. And the results are super cute, super interesting. And this is still a proof of concept, okay? The Fed is not going to move their models to a quantum computer in 2023. But I really believe that there is a good chance that, in 2040, a lot of economics will be done with quantum computers. Hopefully, then, someone will remember that I wrote the first paper on this.

Beckworth: Well, that’s interesting. I was thinking, you sounded a lot like my professors in grad school who would talk about, "Back in the good old days, when you had to sign up to use a mainframe computer at the university, and there's only one big mainframe computer. And if you're lucky you'd get a slot." So, you found this quantum annealer computer, and you were able to tap into it. But what I did get out of your abstract, and reading through your paper, is that it really makes much more efficient calculations, and it takes just far less time to solve the model, which would make all of our lives more efficient, more productive.

Fernandez-Villaverde: Exactly. In a class of models, not in all models, I want to be clear about it. But there is a very interesting class of problems that involves a lot of economics where quantum hardware really, really promises amazing advantages. But again, this is not going to be December, 2023. This is going to be December, 2040. We are the crazy dude on campus that in 1950 shows up and says, "I heard about something called a computer." And everyone was like, "Ah, whatever."

Beckworth: Well, you know what this means, though, Jesus, this is going to be a highly cited paper. You might get a Nobel prize, if you're still alive 30 years from now, because you'll be the first one to do macroeconomics on a quantum computer. Okay, well, let's talk to more germane AI, if I may. And I want to talk about several of your articles. I don't think we have time to do all of them, but one issue I want to touch on is, with AI, there's been this uptick in interest, "Well, maybe now with AI and big data, we can solve the knowledge problem. We can have digital socialism. We don't need markets anymore." And you've written several pieces on this, and we'll provide links in the show notes. But for the sake of time, maybe summarize, what is the knowledge problem, and why do you think AI does not solve it?

Why AI Will Not Solve the Knowledge Problem

Fernandez-Villaverde: So, the idea of the knowledge problem, it has been proposed by different economists. This goes back to the old socialist calculation debate, Ludwig Von Mises, and FA Hayek. And basically, the argument, and I think the one who articulated more clearly was FA Hayek in a very famous paper in 1944, which is actually one of the papers that the Nobel Committee cited when he got an award at the Nobel Prize, is that the problem of central planning is not that, given the information available, you find the optimal solution. The problem is how you get that information, okay? And the analogy I give in the paper is my own experience. I was Director of Graduate Studies at the Department of Economics at Penn for six years, and I had to allocate graduate students to TA-ships to teaching assistant-ships, okay?

Fernandez-Villaverde: Computationally, that problem is quite simple. I don't need a super computer. The problem was the knowledge problem of getting the preferences of graduate students about which classes they wanted to be TA [for], getting the information about their abilities, and providing them with the right incentives to do the right thing on being good TAs. And why this problem is very difficult is because everyone has an incentive to lie. It's much better to be the teaching assistant for a small class with 20 students of something advanced than being a teaching assistant for a large introduction to economics class. And yet, you probably want to put the best teachers to the introduction to economics where they are going to affect, or they are going to have an impact, of 500 students than not on the small class where you have 20.

Fernandez-Villaverde: My problem was never to open the Excel file and assign students to classes. That's computationally very simple. It's just, how good are you as a TA, do I provide you with the right incentives to be a right TA? And that's the knowledge problem. And what I try to argue is that, machine learning doesn't get around this problem, because machine learning is about solving the optimization problem once you have all the data, but it's never about getting the data. Or for instance, another very important aspect of the knowledge problem is the so-called tacit knowledge problem. A lot of what we know is impossible to verbalize.

Fernandez-Villaverde: Let me give you an anecdote. My mom was very good with clothes. This European lady is very elegant, really likes clothes. And my wife could never understand when my mom [would] say, "Oh, that's a nice shirt," or "That's not a nice shirt." My wife, she's American, she tried once to treat my mom. She bought three white shirts, and put [them] on a table, and asked my mom, run them by price. Of course, my mom got it perfectly right. And my wife was asking, "Why? How can you tell?" And my mom was, "I don't know, but you can tell." And I think that anyone who has experience in the fashion industry, who has experience in a lot of things, knows that there is a lot of things we know, and yet we cannot explain them. How is machine learning ever going to get around that tacit knowledge problem? Maybe, I don't know. It's advancing so fast. But my strong prior is that machine learning will never get around that problem, or at least in a sufficient, developed way to really get around the central planning problem, the socialist calculation challenge.

Beckworth: Yeah. So let me just reiterate what you said there, the market process itself is a discovery process, and it's revealing preferences, as well as, I would say, productive capacity. So, it's true. We don't know from one hour to the next what we want to do. Our preferences may change. And the way that's revealed in the marketplace is through prices, through transactions. So, it is hard to know all of the individual preferences out there, and you need some kind of system that can incentivize the revelation of them. The flip side is also productive capacity. We don't really know what potential GDP is, but we have markets doing their best to bring it all together. And that's never going to go away. And I think that's something, as you said, people don't appreciate. But your second point, I think, is really powerful, too. Machine learning takes existing revealed data. It doesn't tease out new data. And I think that's another important insight. And I would add a final point to all of those, is that, who's to say with AI we don't even get more complicated economic structures, right? With even more hidden preferences and particular capacities, the economic calculus problem becomes even tougher with AI.

Fernandez-Villaverde: Okay. Let me give you another anecdote. In my computer, on Google, you can select your language. So you can select… When you open Google Chrome, you can do it in Spanish or in English. On my operating system, I picked Spanish at the beginning. And that was a bad idea. And the reason that was a bad idea was because, of course Google is very smart. So, when they show me advertisements and they show me content, they say, "Okay. You are a Spanish speaker in Philadelphia. Chances are that with a 99% probability you're Mexican.” So, it provided me with a lot of suggestions and content Mexican-related, and you know, God bless Mexicans. I don't have any problem with them. I'm just not that interested in Mexican TV shows. But, some of your listeners may know that, in Spain, there are a few other official languages. And one of them is called Galician, [which is] only spoken in Spanish. So, I switched my Google to Galician, and now all of the content that I get offered by Google is from Spain, which is much more relevant for my taste.

Beckworth: Yeah.

Fernandez-Villaverde: Right there, you have an example where Google is trying to use artificial intelligence, and I'm trying to outsmart Google by lying about the language in which I prefer my operating system.

Beckworth: Okay, one last question on AI. I want to tie this back into our demographic discussion. We're having a decline in labor supply, just fewer physical bodies. We're going to maybe have fewer ideas generated as well. All of those things suggest a lower GDP growth path in the future, but there's always that leftover term, total productivity, that residual, and maybe AI could change all of that. Maybe that's our hope. Automation, robots, AI. But you didn't seem too optimistic in your papers. Tell us why.

AI and Shrinking GDP Growth

Fernandez-Villaverde: Well, two reasons. One is just about the space for economic growth. What do I mean by that? A lot of our GDP is not about producing TVs. We have gotten so good at producing TVs and computers and cars that, even if we got much better, that's not going to provide a lot of economic growth. Really, where we spend our money now is in restaurant services, is in health, is in education. And maybe artificial intelligence can help a little bit at having a better restaurant experience. I just don't think there is that much scope for having a better restaurant experience thanks to artificial intelligence. And then, with respect to things like health, I think that people, coming back to my point at the very beginning of the interview, do not understand national income and product accounts. What do I mean by that? Think about the way in which we record health GDP. It's basically cost. So funny enough, if we make health cheaper because we have artificial intelligence in [inaudible], in GDP, that will show up as lower GDP.

Fernandez-Villaverde: Now, welfare is going to be higher, so I'm not denying that. But if it takes me $100 to do a medical diagnosis, then now it takes $2,000, according to GDP, we are actually going to lose GDP. Now, that may be an argument to change the definition of GDP. And I fully agree that we need to start rethinking GDP. It's just that the way we compute things, if you are thinking that GDP is going to grow thanks to the fact that we have artificial intelligence, in many sectors, no, it's going to go down. This is the same argument with ATMs. We used to have a teller at the bank. That's GDP. Now, we open an ATM or just an electronic wallet, that's not GDP.

Beckworth: Very interesting. Well, Jesus, our time has come to an end, but it's good to have the first quantum macroeconomist on the podcast. And we appreciate you coming back on the show.

Fernandez-Villaverde: Thank you.

Photo by Richard A. Brooks via Getty Images