Antonio Fatás is a professor of economics at INSEAD, an international business school with campuses in Singapore, France, and Abu Dhabi. Antonio joins David on Macro Musings to talk about hysteresis and the business cycle. Specifically, David and Antonio discuss the history of the academic literature on business cycle and trend, the impact of the Kydland and Prescott model, and how endogenous growth models play into hysteresis.
Support Macro Musings and get a free NDGP Targeting mug.
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: Antonio, welcome to the show.
Antonio Fatás: Thank you.
Beckworth: Oh, it's great to have you on. And I've read your work over the years and you've written a lot of interesting papers. And the area today we want to focus in on is your work on hysteresis and the Business Cycle. But to be clear to listeners, you have lots of other interesting work as well. And hopefully today, we'll be able to talk about some of that, particularly the digital currency work you've done. Before we get into this, though, in your interesting paper that you've done on hysteresis and the business cycle, tell the listeners a little bit about yourself. How did you get into economics?
Fatás: I mean, I grew up in a country that when I was young, there was a dictator, so conversations on politics were everywhere at the dinner table at my house. So I became very aware of political conversations, social debates when I was growing up. At some point when I had to decide, when I went to university, what to study, I thought economics was a good subject. I was really hoping that economics would be a way for me to think about some of those economic and social issues, maybe one day get involved in things like research or policy making. So that was the beginning. And once I studied economics, then I liked it. And of course, I followed with a PhD and then I went into an academic career.
Beckworth: Yeah, I know you've done work with some pretty prominent people yourself. You've done work with Larry Summers. Is that right? And others. So how did you guys connect?
Fatás: We knew each other from my PhD times. And I always of course appreciated his work. And he had seen, in fact, a blog post that I had written on issues which are related to what we're discussing today on hysteresis, it was purely just a chart that I thought was interesting to write on a blog. And he saw it and we started a conversation by email. And obviously, he was someone who was very interested in issues around hysteresis. He has written lots of great things about it, and I had written a lot of it as well over the last years. And we came up with an idea for the research paper. And that's how he came up.
Beckworth: Okay. Well, how did you get into hysteresis? We'll talk about this more later, but hysteresis really isn't the dominant view or hasn't been the dominant view in the academy. So how did you find your way into it?
Fatás: I got there because when I was writing my PhD... This was long time ago, early in the 1990s, there had been obviously a wave of research on endogenous growth that been happening maybe over the previous 10 decades. And the moment you make growth endogenous, it just seems to be an obvious proposition to try to think about how the cycle affects endogenous growth. So to me, it was such a natural question. And it was a question that, as you said, had not been asked enough.
Fatás: So during my dissertation, to me, it was obvious that someone had to write about hysteresis. There were other people around that time having the same ideas. I think there was a wave of papers in the early '90s that naturally connected endogenous growth theory to the business cycle. And what I was doing was one of those papers. I think then the theory died out a little bit. I think there was that wave that was natural. And only, I think recently, we've seen a lot more interest in the subject again.
Beckworth: Yeah, very fascinating. Reading your paper, it never dawned on me the link between endogenous growth theory and hysteresis till I read your paper. But man, it seems so obvious now looking back. And we'll come back later to talk about it. But it's neat to hear that doing your dissertation, that was a eureka moment. You're like, "Hey, wait a minute. Endogenous growth theory is more than just R&D and competition policy. It could go from cycles to trend even though that wasn't the focus." That's fascinating to hear your story. So, I wanted to bring you on today, Antonio, because there's been a lot of discussion that's related to hysteresis, maybe some of it's implicit, but there's been a lot of discussion here in the United States about overheating.
What is Hysterisis?
Fatás: As you know, there've been a number of pandemic relief bills that have been passed, almost $5 trillion, 20% of GDP. The last one, $1.9 trillion passed by the Biden administration really generated this firestorm of conversation, debate. People that were, frankly, surprising to me to hear them get worked up about it, Larry Summers, and some others too, Olivier Blanchard, big names. They were worried we were putting too much heat into the economy. It would dip higher than potential GDP. And then others were like, "No, we've got to do this. We haven't returned to full employment." And I think... Maybe I haven't heard many people say this but I think one of the implicit arguments for those supporting this is that reverse hysteresis argument that even if we go past the official CBO's potential output level, maybe there's actually a latent potential output level.
Beckworth: We don't really observe or even estimate, and we can actually push that up. And we'll come back to these ideas in a minute. But I think it's a very topical discussion right now. This is really at the center of what's being tried. It's an experiment. Lots of people are talking about how the policy paradigm is shifting towards running the economy hot versus being more conservative and dialing it down a few notches. And let's begin maybe just with a definition of hysteresis, a simple definition before we get into your paper. But how would you define hysteresis?
Fatás: Hysteresis, if you define it using the standard definition in many other sciences, not just economics, it says that the state of the economy today depends on its history. That's the standard definition. Now, if you take an economic angle and you think about cycles and trend, it says, the position of the economy in the long-term... So if I want to do a forecast of GDP in the long-term, it's going to depend on the path of the economy that we're going to follow from now until then. And that naturally means that whatever happens during the cycle, how fast the recovery is, how slow the recovery is, will permanently have an effect on GDP.
If I want to do a forecast of GDP in the long-term, it's going to depend on the path of the economy that we're going to follow from now until then. And that naturally means that whatever happens during the cycle, how fast the recovery is, how slow the recovery is, will permanently have an effect on GDP.
Beckworth: Okay. And as we'll discuss, this hasn't been the standard dominant view for some time. There's been a clean separation between the business cycle and trend growth. Before we get to that, and again get into your paper, just one more question. I'm going to throw to you an analogy that I often use to describe hysteresis. And it's the one that uses athletics or sports. Since you're the expert here, you can correct me or maybe fix my analogy, make it more proper. But here's the way I think about it. Say that you, Antonio, can bench-press 400 pounds, all right? And you're an amazing athlete. You're in great shape, and then you get sick. That's the recession, you get sick and you're sick for six months. Maybe you got COVID... Let's say you got COVID to make it very topical.
Beckworth: And as a result, after you recover, you go back to the gym and lo and behold, you can only bench-press 200 pounds. That's now your potential. That's as much as you can do. Now, you could go to the gym and you could work out and settle for 200 pounds bench press, and maybe you'd be fit and feel good and stuff. But maybe if you go to the gym and you push yourself really hard, you add heat, you overheat, you work really hard, you can gain that extra strength and get back up to 400 pounds. You could say 200 is your max. Or you could say, "Hey, I want to get back up to 400." And that's how I think about hysteresis that we go through a real severe recession, potential GDP falls permanently, and we can settle for that or we can aim to get back up to that latent unobserved value. Is that a decent analogy or not?
Fatás: It's a good analogy. But your analogy being something which is very much present in the academic literature, which is, you've done an analogy in levels. So either I bench press 400 or 200. So if I don't work hard enough after COVID, I'm going to get stuck at 200. Now, there's a lot of academic work on multiple equilibria. And that's what you just said implicitly in your bench press example, right?
Fatás: I got stuck at 200. And I think that's a good way to think about intuition of hysteresis but I personally feel more comfortable with the stories that have growth somewhere in there. So if I can replace your analogy with someone who is -
Beckworth: Yes, please do.
Fatás: ... similar, I start bench pressing 100 and if I keep working hard, every year I add 10 to that. It's 110, 120, 130. Of course, one day I'm going to reach my maximum but let's not think about that maximum yet. Now, I get sick and I stop training for a year. So I drop 10 of that improvement, right? Because, I didn't train enough that year. Now, is there a way for me to train harder the year after to get caught up with those 10 that I lost? And that I think is a more natural analogy when you truly think about growth in an economy, which is when you go into a crisis, we stop doing what drives the trend. We stop growing, not just because of the crisis but we fundamentally stop the trend from growing. Now, if we go back to normal next year and we don't do that extra effort to recover what we lost in the year that we were in a recession, yeah, we'll keep improving but we'll always be in this parallel line that never matches the pre-crisis trend or, in your analogy, the pre-COVID trend.
Beckworth: Yeah. You're always going to be permanently 10 pounds below where you could be. And I like the way you've changed it because the difference between someone bench-pressing and the economy is someone's going to die at some point. There's literally a limit you're going to hit as a human, but the economy is going to grow forever. So that potential is going to grow forever. So the incremental 10 pounds, I think, is a better analogy. So yeah, I like that and I think it's helpful and useful to think through that as we go through this. Okay. So let's move to your paper. It's titled “Hysteresis and the Business Cycle”. You co-authored it. And we'll provide a link on the show page to it so listeners can go read it. And I really encourage listeners to check it out.
Beckworth: I learned a lot from this. Also, it was neat, Antonio, because it took me back to graduate school, particularly my time series classes. I was reading through all the stuff on unit roots and stationarity and stuff, great memories there. But it's also... I think It's a great synthesis of everything that's been discussed in this literature. And let's begin with a discussion of cycle versus trend. And this interestingly goes all the way back to Burns and Mitchell in 1946 when they first started looking at the business cycle. So walk us through the history of our understanding of cycle and trend.
Understanding the Business Cycle and Trend
Fatás: And I think if you go back in history, of course, we used to write these things in a less formal way than we do today but there was always this understanding that there was something that was of the natural state of the economy. We can call it full employment. We can call it the natural rate of unemployment. We can call it potential output. And business cycles were fluctuations around it. And that's what Burns and Mitchell did. They did it more from a statistical point of view that, "Let's try to characterize these cycles. Let's try to look at these turning points." But in their writings, you always see this reference to this natural state of the economy. I think that's the beginning of the way we think about business cycles. We gravitate towards some natural state, some equilibrium, but we deviate through cycles.
Fatás: There was no real growth in that story. It's just that natural state is whatever it is. It might be growing. Again, very little formalization of these concepts. And I think that's the way we've thought about business cycles and trends for many years after that, is, yes, there is a natural equilibrium on the cycles deviations permit. Now, there was a very interesting question, which is, are deviations only from below or are they symmetric around this natural state? That's a conversation which is not purely about hysteresis but I think, in my view, there's an interesting connection between the two. We can discuss it later more in detail because I think when you combine the two, then you truly change the paradigm of cycles. That asymmetry has been around forever.
I think that's the way we've thought about business cycles and trends for many years after that, is, yes, there is a natural equilibrium on the cycles deviations permit. Now, there was a very interesting question, which is, are deviations only from below or are they symmetric around this natural state?
Fatás: In fact, the NBR cycle definition is asymmetric in nature because there's only negative shock. There's only recessions. Expansions are just normal. Even if we don't want to admit it as academics, when we talk about business cycles and we use those dates, implicitly we have an asymmetric view of business cycle, which takes us back to the plucking model of Friedman, which again, these days is also a very popular one. I do believe there is a very strong interaction between an asymmetric view of the business cycle and the concept of hysteresis.
Beckworth: Yeah, that was interesting going back and reading and thinking about Friedman because the plucking model was in 1964 as I read in your paper. And he had his famous American Economic Association speech as president in 1968, which defined for this whole generation the natural rate hypothesis. There's this natural trend and we're deviating around it, and yet you have a very different paper in 1964. Have you ever thought about what these two different faces of Milton Friedman? Maybe, they're not opposed but it's just interesting how he has two very different views of the business cycle.
Fatás: You can have a view of a natural rate and you still believe in asymmetric cycles. I think the big question is, what constitutes a cycle? What causes it? Are there positive or negative shock? Are those things symmetric? And I think we all understand when we tried to write formal models, symmetric models are beautiful. They're easy to write, they're easy to solve. So it's very easy for us to think about symmetric shocks. When we, as academics, are invited to talk in a podcast like this one or to give an interview with the press, all of us, I think, switch to an asymmetric view of the business cycle, all of us.
Fatás: We like to talk about the negative shocks, and so rarely we talk about these shocks that cause a positive cycle. We very rarely talk about moments where unemployment is clearly below its natural rate. Yeah, we talk about moments of overheating, but those are really, really special events. I think that's the reality of our research and I think it shows in Friedman's writing, which he felt it was very natural to think about asymmetric cycles. But as an academic, he also felt that the notion of a natural state was also a very natural way to think about the economy. But I think the two can be combined. There can be a natural state. It just happens that we only have negative shocks.
We like to talk about the negative shocks, and so rarely we talk about these shocks that cause a positive cycle. We very rarely talk about moments where unemployment is clearly below its natural rate. Yeah, we talk about moments of overheating, but those are really, really special events.
Beckworth: That's fair. Yeah. The other thing about Friedman that you highlight in his paper... Milton Friedman, Edmund Phelps at the same time had these talks, these speeches that were very influential in terms of the natural rate hypothesis, but they occurred when they did for a reason. They occurred in the context of the great inflation. So it's interesting to see all the implications or ramifications of the great inflation. They had a lot of long-lasting effects.
Beckworth: I was just talking to somebody how the great inflation led to the creation of money market funds, which are still with us today. But here we have Friedman responding to the great inflation. And as you note in your paper, the great inflation really forced people to think about that trend growth and separating it from the cyclical part. And you go from Friedman and Phelps and a few others, and you go to the next, I think, pivotal development in this story is Kydland and Prescott's 1982 paper on the Real Business Cycle. Talk us through that. What did that paper do and what trajectory did it put us on?
Impact of the Kydland and Prescott Model
Fatás: I think it put us on a trajectory that deviated us a lot from thinking about the issues of hysteresis as an asymmetric cycle. First, it integrated growth and business cycles but coming from the other side. It said, "Look, the trend is, in fact, a stochastic. So growth itself is a stochastic. And there could be shocks to technology which are positive or negative. So some years, we're luckier and we innovate faster. Some years, we innovate at a slower rate. And in fact, these are the shocks that are going to drive the cycle.
Fatás: Now, I think what happened is that the formality of this model and the beauty of this model, from an academic point of view, it really took off at that point. It was a very strong formalization of ideas that were very neat and as you know well, this has dominated a lot of hard research. Even today, people who write papers that have a Keynesian view of the world, they typically start with a model that resembles a real business cycle model. So again, I think that created a paradigm that, from a formalization point of view, dominated anything else that was there.
Fatás: From my point of view, I think [the Kydland and Prescott model] was a little bit of a detour that focused our attention on the wrong side. It didn't let us think carefully about the interactions between the business cycle and growth. Also, it didn't let us think carefully about asymmetric business cycles, because any model of asymmetric business cycles cannot look as nice and neat as a standard symmetric model of the business cycle like the standard Kydland and Prescott model.
From my point of view, I think [the Kydland and Prescott model] was a little bit of a detour that focused our attention on the wrong side. It didn't let us think carefully about the interactions between the business cycle and growth. Also, it didn't let us think carefully about asymmetric business cycles.
Beckworth: Yeah. What's interesting is that model is still with us today, as you mentioned. They're the workhorse dominant model, the business cycle, the new Keynesian model has embedded in it this that the long run growth part is a real business cycle. And then they have some shocks that work themselves out over time. There's no interaction. This is akin to the classical dichotomy that in the long run, the supply side and this demand side will separate the too. And it really came to be a very forceful idea. And I think... Correct me if I'm wrong, classical dichotomy goes way back historically but this is the modern manifestation of it, is that right?
Fatás: It is. Again, it goes back to the discussion we had earlier that if you believe there's a natural state of the economy and we always gravitate there, you can think about that the state of the economy is determined by the supply side and then there's cycles. And then some people believe that demand matters during those cycles. So counter-cyclical policies are useful to reduce the size, the amplitude of the cycles. But they can never touch that natural state, which is driven by the supply side. Again, this is another manifestation on the same concept that you can truly separate what is driven by supply and we typically associate that with the long-term and what drives the cycle and typically think about the possibility that demand can influence the cycle, therefore, counter-cyclical policies can be effective.
Beckworth: Yeah. I think, just as another example of this, you look in many undergraduate textbooks in macroeconomics... If you've looked for an aggregate demand or the supply model or something similar to that, what you will see is an aggregate demand curve. You might see a short run aggregate supply curve or a Phillips curve, but then you'll see a long run aggregate supply curve or a long run solo growth curve in there, which is... It's straight up and down. There's no relation between these other variables and it's still very pervasive in the literature. Now, something else you highlighted, it's influence really affected everything we did.
Beckworth: Again, as I read this, memories flashed back to some of this literature. But you mentioned Olivier Blanchard and Danny Quah, 1989, they had a paper that... I've used their identification method many times in my own research, but they have this empirical paper where they identify supply shocks, demand shocks. And demand shocks can only temporarily affect supply. Supply shock can have a permanent effect, both on demand and supply. Again, this classical dichotomy... And it worked... I'll tell you, if you use it and you plug in money, you can identify a liquidity effect. I always was fascinated by it and I used it. But over time, people have raised questions, is that an appropriate separation in the data? But it's been very influential, that framing.
Fatás: Yeah. It has been very influential on... I think the reason is exactly how nice and how neat it is, as you said. We've all used it. I remember the first time I'd read that paper, I thought, "This is brilliant. This is such a nice way to think about supply and demand shocks." We've all struggled thinking about how we identify them, here is a paper that tells us how to identify them. But again, fundamentally, if you don't believe the paradigm that can separate supply and demand shocks that that is a flawed method.
Fatás: I mean, it's based on an assumption, thus the identifying assumption that supply is independent from demand. And if that assumption is wrong, of course, the outcome of that estimation, the outcome of that identification is going to be wrong. But again, we go back to this notion that simplicity, symmetry really pays off in academia. And I think some of those developments that you saw during these years, I think they were driven by the beauty of these models and the simplicity of these methods that all of us academics appreciated. And we all use them.
We go back to this notion that simplicity, symmetry really pays off in academia. And I think some of those developments that you saw during these years, I think they were driven by the beauty of these models and the simplicity of these methods that all of us academics appreciated. And we all use them.
Beckworth: Yeah. Okay. There's the appeal of them due to their simplicity, the beauty, the nice and wonderful impulse response functions you see. Let me throw another suggestion out there. Maybe it was also a product of its time. So Blanchard and Quah, 1989, so this is during the Great Moderation. I wonder to what extent the Great Moderation led us astray, this stable period, at least it appeared stable at the time, and it fostered a desire to buy into this classical dichotomy. I wonder as a comparison this past decade, you hear more discussion about hysteresis when you look at the slow recovery in the past decade. But if you go back to the great moderation, I wonder to what extent this view was made easier to accept because of the macro environment it was in.
Fatás: My sense, I think it goes back to the 1970s more than the 1980s. I think, as you said, in the 1970s, we economists, we took a role to tell the rest of society that demand policies have a limit. Again, there was the high inflation and there was always people who thought, "Well, the government can fix this by spending more. The central bank can fix this by increasing liquidity." And we economists, we felt very powerful telling people, "No, it doesn't work. There's a natural rate. You cannot go beyond a natural rate." Then we found out a usefulness in telling people that there was a limit to demand policies.
Fatás: So I think that carried over through the eighties. In the eighties, from a political point of view, there's the conservative revolution that says, "Yeah, be careful, governments don't have a strong role. And that's what happened. Now, you are right. It was also the great moderation at least in the U.S. Not true in other countries, in the '80s, things were okay. There was a long expansion. Of course, the early '80s crisis was painful but there's a very strong recovery from the early '80s crisis.
Fatás: The notion of hysteresis, the notion that the supply side could be affected by cycles was not too visible in the data. Of course, what is interesting is one of the same authors of that paper, Olivier Blanchard, at the same time was writing a paper about hysteresis on the other side of the Atlantic because in Europe, the cycles did not look like the U.S. And in particular, the labor market was being hummer by the shocks of the 1970s and the '80s.
Fatás: Again, there the data was asking for a model about hysteresis. Now, it wasn't full hysteresis, the way I define it, but it was hysteresis in the labor market because the unemployment rate would not go back to normal. Again, you can see a little bit of an author like Olivier Blanchard is struggling, "Okay. How do I think about the natural rate?" Depending on which continent you looked at, you have a very different view of the natural rate.
Beckworth: Well, Olivier Blanchard, he's more international than most Americans. I think I'll put myself in this camp. I think as an American, we tend to be very inward-looking. Great observation that most people know who's getting elected in the U.S. But most Americans have no idea who's getting elected overseas. We're very focused on America. I think that's true also as a macro economist. If you're an American macroeconomist, you look at the U.S... I mean, we were told as youngsters, we can approximate a closed economy U.S. Ignore the rest of the world. I think it's lulled us, made us complacent just to take for granted this framework. And I think again, the last decade has been a good check, a humbling experience for many to say, "Well, maybe it's more complicated than that."
Beckworth: Okay. So we've covered these models coming out of the '70s that really put us on a path to separate supply over the long run or the trend path from the cycle. But also in the 1980s, something emerges, the endogenous growth theory. And as I mentioned before, I just hadn't made this connection. So I think of Paul Romer when I think endogenous growth models. I think of his calls for policies that will develop human capital, innovation and everything from R&D, science policy to more competition, even population growth is important... These things are what I think of in endogenous growth theory. But you opened up a whole new world for me to think about this from hysteresis. So maybe you can walk us through these models and then tie it into the idea of hysteresis.
Endogenous Growth Model and Hysteresis
Fatás: This was the time when I was writing my PhD dissertation. I was thinking about business cycles. That was my main focus of my dissertation. But then as you start looking at these endogenous growth models, you realize that mechanically almost by default, if you have a model where growth is endogenous and you believe that whatever drives growth, whether it's R&D, whether it's investment, whether it is education, whether it is innovation, whatever it is, if that phenomenon that drives growth gets affected by the business cycle, then you're going to have the trend slowing down during a recession. So when the recession is over, you never go back to the same level. And this is just standard. You don't need to think about any complicated logic, you just need to put the two together. That to me was a little bit of a revelation, which is not rocket science.
Fatás: If you combine these two streams of research, business cycles and endogenous growth, then we have a new paradigm here. So that's how I started thinking about these two things together in a very natural way. I didn't feel it took me that much to think about that. What was harder of course is to come up with a model that was simple enough, that was convincing enough, and that was matching the stylized fact of the literature. And here is where I struggle but I think the literature on hysteresis has struggle, which is, we all love the endogenous growth literature. It sounds very insightful. It helps us think about growth, competition, R&D. At the same time, there's nothing that I could call a workhorse model of endogenous growth that produces a standard way to think about endogenous growth for the U.S. or for any other country.
Fatás: Again, if you ask me, "How would the U.S. growth change if R&D increased by one percentage point?" I don't think we have a model that we will all agree that would produce that estimate. Now, we assume R&D helps. We assume investment helps. We assume human capital helps. And we have a lot of those cross-country models with convergence that help us think a little bit about growth, but we stopped doing cross-country progressions a few years ago because we thought, "Oh, they're not good enough for our econometric understanding." So we stop thinking too much about what are exactly the parameters and the right channels for endogenous growth to work. And without that, it's very hard to write a model of hysteresis that people say, "I get it. This is how I'm going to calibrate now the next shock on the next shock."
Fatás: So I think that's the struggle that I've had in this literature, my own struggle but also the literature, which is there's a natural connection between the two, but we need to admit that we do not have an accepted model of endogenous growth the same way RBC became the standard model of business cycles. That we all accept. And if I'd write an RBC model, no one is going to challenge me. But with endogenous growth, it's very much paper by paper. I'm going to write a paper using this particular channel. You're going to write a paper using another channel. And I think it is much harder to advance the literature when that happens than when we all using the same.
We need to admit that we do not have an accepted model of endogenous growth the same way RBC became the standard model of business cycles. That we all accept. And if I'd write an RBC model, no one is going to challenge me. But with endogenous growth, it's very much paper by paper.
Beckworth: That's interesting. Let's go ahead and talk about the models since you brought them up, the theory for endogenous growth. And you mentioned, there's three main models, so you can take various forms of them. But you mentioned there's a learning by doing model, the Romer AK model and then R&D model. So maybe just briefly summarize those for our listeners.
Three Kinds of Endogenous Growth Models
Fatás: And so there's three ways... There are many ways, but I think three main ways to think about endogenous growth. One is, the learning by doing, which is the notion that when you have more activity, people are more engaged into production. You learn how to innovate faster or you learn how to produce at a more efficient level. So the more activity there is, the more growth you're going to see during the same time and thus the learning by doing. The AK model, which refers to the production function that Romer used says...
Fatás: It's all about investment. Growth in the long run is about investment in physical capital and under some assumptions about the production function that this constant returns to scale on physical capital. You're going to generate permanent growth by investing. And the R&D model says, "Well, growth is about innovation. Innovation is driven by the time and the resources we spend into research. And that's the engine of growth."
Fatás: Now, I think the three things are complimentary. I wouldn't think about them as exclusive models. They're three channels through which one can model the notion that, over time, productivity increases and increases through forces that have to do with learning by doing, with investment in physical capital or investment in R &D. Of course, human capital comes also in any of these three channels as well, right?
Beckworth: Yeah. Just to make this concrete, so if we go back over the past decade, for example, we had a slow recovery coming out of the great recession, U.S. Europe, other places. What these theories would say is that, "Man, there's a whole lot of human capital formation that didn't take place." In fact, I think we can look at the end of this past decade when we saw... The economy was never running really, really hot but it did continue for a long, long time. The expansion continued and that by 2019, what we saw were people coming off of, for example, disability payroll. So people who were unemployed because of disability, they were entering the labor force.
Beckworth: And you see this healing nature. And many people we're saying that those folks are structurally unemployed. They're not going to get back in but low and behold, they did. And I think the argument from these theories, and correct me if I'm wrong, is that look, you would have a whole lot more of that, people learning by doing on the job, you'd have firms kept more capital formation, smarter technologies if you were just more engaged, if you were running the economy a little bit hotter than we did, there's actually gains to be made that weren't recognized. Is that a fair real-world application?
Fatás: Yeah, I think it's a perfect one. And as you go through some of the things you said, some are easy to model and quantify. Some are very hard. So the easy ones are when you think about workers, right? Again, those workers that were unemployed in the last expansion or out of the labor force for a decade. And they only showed up in the last moment in 2019. If you do a counterfactual, what if they had been employed from day one, since 2009? You would think that they were probably more efficient by now because of learning by doing. That's an easy one.
Fatás: And we can also think in the other direction, some of the people who got fired in 2008, maybe never came back to the labor force. So if they hadn't been fired, they would have been in the labor force. We would have had more production, slightly more growth. When it becomes really hard is when you go beyond that human capital and the skills argument, and you start talking about physical capital investment. You thinking about innovation, why? Because then you ask the question, so firms were innovating our investment in the whole decade, since 2009 and 2019.
Fatás: Now, what type of cyclical conditions would have maximized the amount of innovation and investment that firms could have done? What type of path for the unemployment rate, what type of path for the output gap would have maximized that innovation? That is really where it gets hard. Because you have to be careful. There is a limit to the amount of innovation we can do. And here we go back to running the economy hot, a high pressure economy. We know that it's not that by keep pushing, we're going to have a growth in the U.S. of 10% of productivity every year. Of course not, no one thinks like this.
Fatás: Measuring that gap, which takes us back to the cycle trend, the composition to the measure of the output gap, that's the power of which I find the hysteresis literature does not have yet an answer because we don't have really a model that says, "Here is a nice way to think about how much more can we push the economy closer to potential and what would be the long-term benefits of that?" The logic is very clear in my mind. The models are very neat. It's very possible to see this connection. But again, estimating a cyclic, how much more growth we can get out of innovation, out of investment, that I find a challenge and I don't think we have an answer yet.
Beckworth: I guess this experiment we're doing now will maybe answer some of those questions over the next few years. We'll see. I'm sure lots of papers will be written. And I'm hopeful that we'll have some productive results. And I think it's fair to question how the money is getting allocated, what it is being spent on. But one of the critiques is how is this going to help productive activity? I've heard one person say, "Oh, you're just doing money illusion. You're going to lower real wages to get people to work" Some simple story, but I think the reverse hysteresis...
Beckworth: This is a reverse hysteresis story is that by running the economy hot, there's going to be all this stuff that we talked about, learning by doing, human capital formation. And we're actually going to see an increase in productivity, which in turn would raise the real wage. The marginal product of each wage is going to go up. Let me rephrase this, the good story that we could tell for the next few years is that the spending is going to raise the protective capacity of the economy and increase real wages, not just some temporary bout of money illusion. Is that a fair assessment of what the best case would be?
Fatás: I think to me, in the current debate, here is what I would like to combine hysteresis with an asymmetric view of the business cycle. Because if you combine the two, then you can make a very strong point. Again, I go back to the previous expansion. It took us 10 years to bring the unemployment rate down to 3.5 in the U.S. Now, that is really not what our economic models predict that you get a shock and it takes a 10-year expansion to go back to normal, to full employment. Why is that not that what our models predict?
Fatás: Because this was the longest expansion ever. If the longest expansion barely took us to full employment, what happened in all the previous ones? Did we ever get there? And this is fundamental question that we, as economists, have not answered yet at least not according to my taste. And this goes back to the plucking model. Do we go back to full employment ever? I'm writing a paper on this right now. We never seen the U.S. a period of low unemployment that persists for years. We've never seen it.
If the longest expansion barely took us to full employment, what happened in all the previous ones? Did we ever get there? And this is fundamental question that we, as economists, have not answered yet at least not according to my taste.
Fatás: Look at the picture of unemployment in the U.S., every time it goes low, it jumps back up. Again, that's not what our models predict. So again, that's the plucking model. There's no hysteresis there. Now, add hysteresis to that and you say, "Well, during those years where unemployment was too high, we were not at potential." If you believe there is hysteresis, we're damaging the long-term potential of the economy little by little. By how much? We don't know yet but certainly there must be some effects on the supplier of the economy. Now, what we're doing this time in the U.S. is let's go as fast as we can towards potential.
Fatás: Now, to me, that sounds like a good idea given that last time it took us 10 years and 10 years sounds suboptimal I think for all of us. And then, yes, let's see how far we can push that potential output. We all know it's very hard to measure potential output. We all know our measures of potential output tends to be incredibly cyclical. In recessions, we all become pessimistic and we immediately lower potential output. But if you have a very strong boom, and in the case of the U.S., I think that's the 1990s, you can keep raising potential output as you go along. So where that line is, I think we need to admit there's a lot of uncertainty.
Fatás: I use the word that you used before, we would experimenting a little bit. I think as an economist, I'm really intrigued to see what is going to happen because I think we're going to try to push that line a little bit farther up. Now, I'm sure of thinking that that's what we're going to see. We're going to see that line moving up. But I think we're going to learn something. I think we're going to come up to a state of affairs in the U.S. where we will be able to test some of these theories in ways that we couldn't test before.
Beckworth: You raised an interesting observation that our data is limited. But that applies to the other side as well. That applies to the standard view of there's the classical dichotomy, the standard macro view in the business cycle models and the new Keynesian workhorse model, they also are relying on just a few observations, the expansions, as you mentioned. The uncertainty goes both ways. I think it's fair to say it goes both ways, not just on the side of those who want to see if reverse hysteresis has merit. I think if we want to be honest, we need to have some humility that both sides really don't have enough data points to reach a conclusion. And it's interesting, you frame this as, and I framed it too just now, as an experiment in reverse hysteresis.
Beckworth: I also like to view this coming, put on my other hat, this is a form of level targeting what we're trying here in the United States, it's makeup policy. And you may know this, I'm a big fan of nominal GDP level targeting. But any level targeting in my book is an improvement over what we've had. Because it empowers us to do rapid make-up policy. And along the way we get to test hysteresis. I think these ideas go hand in hand and I think the average inflation target is a watered down version of a price level target that's going to allow us to try it out.
Beckworth: The Feds have been very clear. They're not going to tighten until they actually see inflation. It's outcome-based versus preemptively going forward. So this will be an experiment. And I also have confidence the Fed is not going to allow inflation to get way out of hand. They'll allow some overshoot, which is what a level target would call for and then we get reverse hysteresis. So I think we'll get a number of questions answered. We'll get questions answered about hysteresis, questions about, how do you do level targeting? Because that whole make-up period, there's a lot of uncertainty.
Beckworth: In fact, I had a conversation with Bennett McCollum who was a great nominal GDP targeting fan. And he had a paper and we exchanged some emails and he was very much a nominal GDP targeter but for growth rates where I was for the level target. And I think one of his big critiques is that period of catch-up growth, there's so much uncertainty, a lot of discretion. It's not very rules-based. And so he had some problems with it.
Beckworth: And I understand that. And I think what we're going through now, I think everyone's learning, how do we respond? The bond market's learning. And I think the Fed's learning. We're all learning. So I think it's a great time to be alive as a macro economist to see these things unfold. We were talking before the show, it's a great time to be alive. I can talk to you in Singapore. Technology... It's a wonderful time to be thinking and talking to interesting people like you about this. Okay. One last question on the model. We talked about the learning by doing model, the Romer AK model, the R&D model. You also mentioned some multiple equilibrium models are part of that package. And I was thinking of Roger Farmer's work. Would you consider him a reverse hysteresis person or not?
What about: Multiple Equilibria and Secular Stagnation
Fatás: What multiple equilibria models have done and they've been around obviously for many decades is those were models that were challenging the notion that there was always a natural state of the economy towards which we come back. Again, that standard notion that we always go back to the same place is challenged when you have more than one equilibria. So in that sense, they freed us, well, with the idea of let's challenge the standard model. And you can make hysteresis in those models that resemble hysteresis. Now, what is a little bit tricky is a lot of those models are in levels. Again, they go back to your analogy on bench-pressing, right? They typically think in levels, which is okay but it's not easy to combine that hysteresis with hysteresis of growth always.
What multiple equilibria models have done...is those were models that were challenging the notion that there was always a natural state of the economy towards which we come back. Again, that standard notion that we always go back to the same place is challenged when you have more than one equilibria.
Fatás: Now, some of those models do think in growth rates and what they do is something which one could even call super hysteresis is in some state of the world that says, "Well, there’s two equilibria, either we grow at 3% or at 1%. And which one we go in is going to depend on many things. It might depend on animal spirits. It might depend on some… whatever the model is." If you jump from three to one, you would have super hysteresis because it's not just that the level gets affected, the growth rate gets affected forever.
Fatás: But if you flip back and forth between one and the other, so you first grow at three, then you spend some time growing at one, then you go back to three, that looks a lot like hysteresis because for a few years you were growing at the wrong rate. And that links to a more statistical literature, which is about growth cycles, which is also a very well-established literature in economics. And we go to models like Hamilton's models, where you tend to think about the state of the economy as growing at high rates and low rates.
Fatás: And that fits naturally into models of hysteresis because that's the story of endogenous growth with a slightly different paradigm. I have the expansion phase where I always grow to something percent. And then I have the recession, where I grow at zero minus one, then I go back to two. That leads to hysteresis because I never ever grow above trend. You see there, the three or the one, nothing higher. It's a natural interaction although these models are very different in nature. They come from a different place. But statistically, you could think that they produce similar outcomes.
Beckworth: Yeah, very interesting. One more question about theory, I said that was the last one. This will be the last one for real. What about Larry Summers' secular stagnation theory? Would that count as a form of hysteresis or not?
Fatás: I mean that hysteresis, I think it has different versions and there are different ways to plug it into academic models. There is a version of that, that almost sounds like a multiple equilibria hysteresis that is stuck in a period of deficient demand. And again, that sounds a lot like the multiple equilibrium models. If you believe we're stuck forever... Again, I wouldn't call it hysteresis, I would call it super hysteresis because you're stuck in a logarithmic equilibrium. But again, if you start going back and forth between different states of the economy, I could feed it into the hysteresis story. So it can be linked but I wouldn't say it's 100% a link to the same type of empirical predictions.
Beckworth: All right. Well, let's move to the evidence. You have a section of your paper that documents the empirical work. You look at the U.S. You look at cross countries, you look at crisis. And let's go to the U.S. So a lot of interesting empirical work was done in the '80s and later. So maybe you can summarize that for us.
The Empirical Evidence
Fatás: When it comes to empirical work, there's an empirical fact that we all accept, I think almost all of us accept, which is fluctuations are persistent. So typically, when there's a shock, when there is a recession, if you look at the long-term effect of that recession, it is somewhere in there or over a long horizon. So we do not go back to a linear trend. So we don't go back to a pre-crisis linear trend.
Fatás: And I think this goes back to the unit root discussions that date back many, many years ago. And I think again, in most time series analysis of U.S. GDP, I think that's what we see. We see a very persistent fluctuation. Now, that matches the notion of hysteresis, that when you have a shock, the shock is not just cyclical, it becomes permanent. But of course it also matches the workhorse model in business cycles, which is the real business cycle model, because in that model shocks are permanent by definition because they affect technology.
Fatás: They affect the parameter in the production function. So then it becomes a little bit of a race. Here, you have two predictions about permanent effects of cyclical shocks. One is about the exogenous component is permanent. The other, the hysteresis story is the endogenous reaction of the economy makes a cyclical shock permanent. So a lot of the empirical literature has been about a fight between these two stories, because I think that persistence of fluctuations is obvious in the U.S. and it's even more obvious outside of the U.S. So again, it doesn't matter which country you look at. The debate is what causes that persistence? Is it supply shocks? Is it the exogenous persistent or is it the hysteresis endogenous reaction to a cyclical shock?
You have two predictions about permanent effects of cyclical shocks. One is about the exogenous component is permanent. The other, the hysteresis story is the endogenous reaction of the economy makes a cyclical shock permanent. So a lot of the empirical literature has been about a fight between these two stories.
Beckworth: I think here is another example where at least for American economists, we've got lulled into complacency. And the great moderation, I think, may have biased our views from all the evidence around the world, and maybe longer time series, because the 1980s, you could plug in the traditional model, supply and demand. They're distinct at long horizons. But as you point out, though, if you were to look around the world, you don't have that clean relationship.
Beckworth: That was maybe a nice artifact of the great moderation, but there's many crises around the world, advanced economies, emerging economies. I think we can say the evidence from the past decade, even in the U.S. now bears this out. But the bottom line is there's unit roots in the data. And I joke on Twitter that we are unit rooting away our prosperity, turned it into a verb. And you just plot out real GDP and you see it drops and then it drops and there's this level effect that tends to happen and is persistent. And what you're saying is this is a common phenomenon throughout the world. Is that fair?
Fatás: That's correct. And it's more common in other countries in the sense that some of these shocks, if you do a standard plot of the linear trend, and then you do a recession and you see how close you go back to the trend, some of these shocks have much bigger effects in other countries than the U.S. Now, why is it that the U.S. looks closer to a linear trend? And I have two hypothesis, which are partly explored in the empirical evidence that we cite in the paper. One is, when you have countries that are growing faster, the engine of growth is very powerful because you always grow faster, in particular emerging markets. Now, if you disturb an engine, which is very powerful, if you disturb a growth rate, which is typically, four doing a recession, so you go from four to zero. If you waste a year of growth, you wasted 4%. Now, when you when you disturb an engine that only grows at two or one and a half, well, you lost about 1.5%.
Beckworth: Great Point.
Fatás: So the permanent effects of cycles should be bigger in countries that grow faster. Now, that's one of my earlier papers in this literature, which is to plot the persistent of fluctuations against the average level of growth. There's an almost perfect correlation. That's one reason that says, "Well, the U.S. because it's an advanced economy, you might not see a lot." Now, the other reason is, "Let's go to another advanced economy, which is Europe, the European countries, or the Euro area, If you want, in particular, in recent crises, you have massive persistent effects." 2008. Very likely 2020, we're going to see the same effects.
Fatás: Now, here we go to the policies. The policies that the US has implemented in particular in the last two cycles have been a lot more aggressive than in Europe, both when it comes to monetary policy and fiscal policy. We're seeing it right now as we speak right now. If my theory is right, if hysteresis is there and cyclical policies matter, that's what you would expect. You would expect permanent effects to be larger in countries where you don't have a strong policies. Now here's the experiment that we running right now, which I find fascinating. If you look at the last OICD forecast, if you look at the last IMF forecast, they both predict that the U.S. economy is going to go back to trend. In fact, it's going to be above trend in a couple of years. Why? Because they attach a lot of value to the current policies.
Fatás: Now, if that happens, that to me would be a very strong proof that hysteresis matters because that forecast is just so unique. We've never made a forecast like that in the middle of a crisis. I mean, not to my knowledge. Now, this is of course, a special crisis. You can say COVID is this special. Yeah. But why is it that the U.S. is the only country where they forecast that? So that, to me, it's going to be a fascinating couple of years ahead if that happens. I think that's going to have to challenge a little bit some of the preconceptions we have about the business cycle.
Beckworth: I can hear someone saying, "Oh, but this is a supply-driven recession. So therefore, it pops back instantly." But your point is very telling, "Well, then why isn't it doing the same thing in other countries? Why isn't it happening in other places?" So you've got to explain that. But the other thing I would mention about this past year, there's been a lot of interesting work. I think that adds to your case for hysteresis, Yvonne warning, Michael Woodford, and the Alp Simsek, Ricardo Cavallaro, they've all done work that shows you really can't have the clean supply shock that isn't intertwined with demand shock. You just really can't have two... In other words, there's no such thing as a pure supply shock. They're definitely going to be linked together. I agree with you.
Beckworth: I think this pandemic should be a lesson for all of us if we look at the totality of the evidence, if we look at why is it we're going to get back to trend, which is so unusual. And maybe that's why it's been hard for some prominent people to embrace it. It's like, "Well, this isn't normal." We just have never tried. Again, we don't have the data points to know what's of the other side of this. But I like what some have said that the risk of going above is less than the risk of staying below for a long time if you look at the total loss in output. All right. Let's turn maybe to some of the policy implications or questions, and let me first go to maybe an analytic question that's related to this.
Beckworth: One of the implications, I think, of what you're saying is, are estimates of potential real GDP or full employment are just way off? We don't really know... We have estimates... In other words, there's almost two measures of potential GDP. There's the one that we think, and then there's this latent unobserved one and maybe even more. If I'm a policy maker, how do I do policy so that I don't go beyond that latent one, the true potential? How do I find my way to the true potential without doing anything more than that?
Fatás: I think when you look at how we estimate potential output, we have many methods, some more sophisticated than others, some using judgment, others just using no judgment at all. In all cases, if you look at the data, these methods end up looking like a standard filter of the data. So you smooth out some of the fluctuations, which in practice means... And this is analysis that I've done myself. If GDP falls by 1%, the typical estimates of potential output, whether it's the CBO in the US or the IMF or the OACD across the world, they typically reduce potential output that same year by about 0.6. So 60% of a fall in GDP is attributed to potential output.
Fatás: Now, that sounds like so pessimistic. So if, as policy maker, I'm using that potential number as my indication of what is the output gap, how much more should I do? I'm really, really becoming very pessimistic about the long-term potential of the economy. Now, I'm not sure policy makers always use those indicators. Maybe in their head, they have a different number which is a slightly less pessimistic. But those are the numbers we see. Those are the numbers that get published.
Fatás: Those are the numbers that influence policy in some countries because of some constraints on policy in particular in Europe, because in Europe, we have a pan-European fiscal framework that says to Italy, that says to Spain, that says to Portugal, watch out, you cannot do a fiscal expansion which is bigger than X. How do they calculate these numbers? Using the estimates of potential output. So here, these estimates enter policy in a direct way. There's very little room for judgment because I'm going to use the formula and we're going to calculate your deficit.
Fatás: I'm going to calculate what potential output is, and I'm going to make sense out of that. So I think we need to be really, really a lot more during crisis about potential output, because otherwise we're all going to undershoot. And the worst thing is if I'm very pessimistic at the beginning of a crisis, and I'm a central bank or a government, and I don't do enough. So I don't push the economy fast enough during the crisis and hysteresis is true and hysteresis is there, then potential output is going to fall.
Fatás: A year later or two, I'm going to say, "I told you so. We should be very pessimistic because look, potential output has indeed fallen." But it has fallen because I didn't do the right thing." And that's my reading of some of the policies we did in the financial crisis in particular, more in Europe than in the U.S. that policies affected potential output. And that pessimism we started with ended up becoming real because we ended up affecting potential output with the wrong policies.
Beckworth: Yeah. Being too tight, itself, can be a determinant of potential GDP. SO moving forward, how does the policymaker know they are running the economy hot? I mean, is what we're doing in the United States a good example of that we have fiscal policy, a lot of direct funding, transfer payments, stimulus checks, unemployment. Big amounts of dollars are going into the economy from fiscal policy. And then the Fed is saying, "Look, we're going to step back." I mean, I think one of the defining features of the Fed's new framework is they're admitting they don't have a good theory of inflation. So they're just going to step back and wait till inflation actually shows itself. And then they'll say, "Okay, now we know we're close." Is that how you run the economy hot or is there some other way you think you should do it?
Fatás: I think that sounds like a good rule of thumb but I think we should be humble and honest that we don't know much about where potential output is and the central bank does not really have a strong theory of inflation. But that means to me that we need to experiment. We need to say, well, if inflation is not there, it must be that we're not very close to potential output. Until we see it, we should not stop. We're stopping too early because maybe inflation overshoots a little bit the target, I think that's a bad idea.
Fatás: Now I'm with you here, the average inflation target framework of the federal reserve, I think, it really fits very well this notion of experimentation because you have a lot more luck to overshoot the 2%. Not only you have luck, maybe you have a commitment to do it. It goes beyond that. And as long as markets understand that, which is a complicated question, but as long as markets understand that, that's the right way of doing it. Again, I think here I'm in favor of experimenting and I do think the U.S. is experimenting right now. And experimentation here is a good word, it's not a bad one. I think that's the right thing to do from a policy point of view.
Beckworth: Okay. Well, that our time is up. Our guest today has been Antonio Fatás. Antonio, thank you so much for coming on the show.
Fatás: Thank you.
Photo by AFP via Getty Images