Joe Gagnon on *25 Years of Excess Unemployment* and the Phillips Curve Debate

The past 25 years of unemployment and inflation data has demonstrated that it might be time to rethink how macroeconomics views the Phillips curve relationship.

Joe Gagnon is a senior fellow at the Peterson Institute for International Economics and was formerly a senior staffer at the Federal Reserve Board of Governors. Joe is also a returning guest to the podcast, and he rejoins Macro Musings to take a look back on the past few years and to discuss his new paper on excess unemployment over the past 25 years. Specifically, David and Joe also discuss the movement of the natural rate of unemployment over time, alternative explanations for the flattening of the Phillips curve, policy implications for the Fed moving forward, 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: Joe, welcome to the show.

Joe Gagnon: Good to be here, David.

Beckworth: It's good to have you back on, and if I'm keeping track correctly, this is your fourth appearance.

Gagnon: Oh, fourth?

Beckworth: Yes, is part of the elite club. I know I haven't gotten you four nominal GDP targeting mugs yet so we're behind on the count, but I know you got at least one.

Gagnon: I do have one.

Beckworth: Yes, yes. So, it is our goal to get them out to everyone in the world, so one day we will all become nominal GDP targeters. But in the meantime, we continue to debate and have fun and think about macroeconomic policy. And Joe, I'm happy to get you back on because we always have a good time chatting up macro. And one thing I wanted to chat with you about is these past few years and lessons learned from these past few years. Where we've been right, where we've been wrong, and I'm going to start this off by doing a mea culpa, admitting a mistake I made. And I've already said this on the show but just to come clean fully. My inflation forecast in early 2021 was just way off, and it's in the paper of record so it's pretty glaring. I had an Op-Ed with Ramesh Ponnuru, my colleague and the title of the piece was, Stop Worrying About Inflation. That was the title, the unfortunate title-

Gagnon: Oh, dear.

Beckworth: ... February 2021. And the reason we got it so wrong is we were relying on market forecasts. So Treasury market, the TIPS, the breakevens. And I bring that up because you actually were warning me Joe, back then. You were like, "David, slow down, bond market isn't always right," and you wrote several interesting pieces that I'd recommend our listeners check out. You made this case that sometimes the bond market gets it right, often it does not. I believe you did a horse race between bond market forecasts, professional forecasters, and households. But in the midst of all that, your big takeaway is: we're not that good at forecasting and have some humility. But I went into the deep end of bond market inflation forecasts and as a result, I have to confess I was terribly wrong there. So I'll start off with that Joe, any thoughts?

Looking Back on Inflation Forecasts

Gagnon: Sure, David. Well, I think it's good that we should have accountability exercises and I'd like to maybe take credit for a couple good calls and maybe a bad call. So the good calls, a long time ago, I think I was calling for more aggressive monetary and fiscal policy in the slow recovery back in 2010, '11, '12. And I think in hindsight, people at the Fed that I talked to think that I was right, that maybe they should have been more aggressive then. But there was a lot of voices against that so…

Gagnon: Then a bad call I made was in 2019, I thought maybe the economy was running a little too hot, not by much but by enough maybe to see a little upward pressure in inflation, and it didn't happen in 2019. And so I got that one wrong, I now think that the equilibrium unemployment rate was even a bit lower than I thought. So, that's mostly a difficulty of measurement issue but I did get that wrong. And then as you say, after COVID, when we had these massive fiscal packages, I just used standard textbook macro to say that this looks inflationary to me. I still can't get my head around the fact that almost no professional forecasters saw this coming, so you were in good company, David.

After COVID, when we had these massive fiscal packages, I just used standard textbook macro to say that this looks inflationary to me. I still can't get my head around the fact that almost no professional forecasters saw this coming

Gagnon: And so I did a series of regression analyses, which you mentioned and saw on our Peterson website showing that neither bond markets nor professional forecasters nor households are good predictors of big changes in inflation. Bond markets react quickly when inflation starts to rise, they catch on quickly but they're not predicting movements. They don't catch turning points, they're very bad at that. But think about COVID, how monetary policy was super easy, QE had been restarted, interest rates were zero, fiscal policy was the biggest deficit since World War II. And yet no one thought that would be inflationary, no one thought that would boost aggregate demand a bit more than the economy could handle. Larry Summers famously pointed that out. Also, Olivier Blanchard and Adam Posen, my colleagues and even Jason Furman also all at PIIE… and PIIE was a hotbed of people who saw the inflation coming so I'm proud of that. But none of us made a point forecast, and I think it's fair to say that what we actually saw was even more than any of us expected. I was thinking something like 4% inflation. And we got 8% so no one got that.

Beckworth: Well, to be fair though, that 4% was based on demand inflation. So maybe you are right if you consider maybe the other 4% is due to supply side factors, give yourself some credit Joe.

Gagnon: I agree with that.

Beckworth: Yeah.

Gagnon: I think that's right. Some of us saw the demand side, Larry Summers saw the demand side, but nobody saw the supply side. The snafus and how that would come in, and of course the Russian war in Ukraine is part of it.

Beckworth: Right. Yeah. So, we want to be careful here. But I want to follow up on what you said, looking back in retrospect, it does seem pretty obvious that high inflation was on the way. The ARP was $1.9 trillion and if you go back and look at the CBOs output gap at the time, it was only $400 billion, early 2021. So just do the simple math, right? You're going to have over a trillion dollars excess. In fact, I like to look at nominal GDP, Joe, as you know, and we have a nominal GDP gap we put up at the Mercatus Center. But even if you just draw a simple trend line from pre-pandemic, you're anywhere from $1.3 to $1.4 trillion larger than you would otherwise have been on.

Beckworth: So the dollar size of the economy is $1.3 to $1.4 trillion larger than it would otherwise had been relative to the pre-pandemic trend. So, that fits well with this $1.9 trillion ARP plan versus the CBOs $400 billion whole at the time in the economy. So the Peterson Institute for International Economics was the voice crying in the wilderness, inflation is coming and we should have paid better attention. So we have looked back at some of the calls we made and you got a number of them right, I got one really wrong. I'll take credit for one call I got right, I was asked to come back to the New York Times after that horrendous Op-Ed. They somehow invited me back, and they asked me to write about who would be the next Fed chair and I said it'll be Jay Powell, he'll keep his job. So I got that one right so-

Gagnon: Okay.

Beckworth: ... I'm batting 50% here. Okay, so let's go on to another potential lesson from these past few years and that is, what did we learn about escaping from the zero lower bound? Because we returned to the zero lower bound and we returned from it back to normal interest rate space. What is your takeaway in terms of, what did it take to escape from the zero lower bound?

Escaping from the Zero Lower Bound

Gagnon: Well, I think it shows that you can escape with a big enough shock, especially if this was a fiscal policy error but it was an inflationary shock. So it showed that escape is possible and I think the question for the future is, are we onto a new fiscal world of large deficits that will keep us on the raise the natural rate of interest indefinitely or not? And that remains to be seen. I do think that the other factors, there's a growing consensus among economists, [that the] long-term equilibrium real interest rate has fallen. And you can debate about over what horizon and how far, but there's a lot of evidence that it's lower than we used to think it was. And people point to a number of things, but the most common ones are demographics.

Gagnon: And if you think about the models, maybe we're near the bottom but it's not going to rise back up for those reasons anytime soon. Maybe 10 or 20 years from now, it will come back up a bit. So, there are the forces at play to keep interest rates low and the question is, is this fiscal world enough to offset that? And I think the jury's out.

There's a growing consensus among economists, [that the] long-term equilibrium real interest rate has fallen. And you can debate about over what horizon and how far, but there's a lot of evidence that it's lower than we used to think it was. And people point to a number of things, but the most common ones are demographics. And if you think about the models, maybe we're near the bottom but it's not going to rise back up for those reasons anytime soon.

Beckworth: Yeah. I mean, fiscal policy was definitely large as we discussed, larger than it was needed. I wonder also to what extent the Fed played a role, the Fed kept rates at zero for a long time. So you had both fiscal policy really strong, monetary policy keeping rates low, looking back longer than they probably should have. And to me that seems pretty obvious that was a powerful combination. I know some might say, "Well, maybe it was just the pure recovery of the economy on its own.” It was just a temporary shutdown, supply side. It reopened, we crashed the economy and it bounced back up really quickly. But it strikes me that large fiscal policy with highly accommodative monetary policy definitely had to play some role in escaping from the zero lower bound. Okay, let's talk about another potential lesson from this period, and that is for the next Fed framework review. And we're going to actually touch on this when you get to your paper that we're going to talk about. But we can jump ahead here, any lessons the Fed should take from these past few years, 2020 up to the present?

Monetary Policy Lessons from the Pandemic Period

Gagnon: Well, from the past two years, the lesson I take… two to three years is... Well, I think COVID is a once in a century shock. So, I'm not sure how much we should worry about lessons from COVID per se. Just because, in our professional life and certainly since economic data we've collected on the basis they are now, we've never seen a shock like this. Maybe you could say the Spanish influenza was but we didn't have the data then, we didn't have unemployment insurance, we were in the gold standard, it was just a different world. So in the modern world, we have no data to compare, and my hope is that this doesn't happen again for another fifty to one hundred years. So I'm not sure how much effort we should spend on understanding that, but some people think it's important. I don't have any lessons. I mean, obviously there were supply chain issues that people didn't foresee, I don't know what to say. I'd say two things though, one thing is I want to talk about in terms of my paper is about the nonlinearity of Phillips curves. I think COVID actually provides more information on the nonlinearity of the inflation process. But let's put that off, we'll talk about that because I think my paper goes back in time and shows there was evidence even before COVID. So, it can all fit in together.

Gagnon: The only thing I'd say is just that, I don't understand why the Fed and virtually all professional forecasters didn't see, as you said, that this massive fiscal stimulus, much, much bigger than any output gap estimate, wouldn't be inflationary. I mean, I still can't get my head around that. So you pointed to bond markets not being concerned and professional forecasters not being concerned so you weren't concerned. So you were in good company, but the bigger question is why weren't professional forecasters worried and why weren't bond markets worried when the standard old fashioned macro textbooks, Keynesians even, if you will, would've predicted a big inflation. I mean, what gives? Why did people not believe in the textbook models? I find that fascinating.

Beckworth: Well, maybe we weren't relying upon them. Even you take a monetarist view of the world, it also would've predicted high inflation too at that time. So, I think Ricardo Reis had a great article or maybe it was a Twitter thread where he went through all the major macro theories. Keynesian, New Keynesian, Monetarism, fiscal theory of the price level. Anyway you cut it, it should have warned you.

Gagnon: Yeah. We're doing QE again, right? So the money supply was expanding. So yeah, Monetarism would say the same thing.

Beckworth: Yeah. Well, here's my takeaway for the Fed and its framework review in 2024, 2025. And listeners will get tired of me saying this, but I think it clearly illustrates at least the supply side challenges in understanding what's driving inflation. So we talked about your 4% forecast for inflation from the demand side's probably true. Then there's this other component, but in real time it's hard to know that, right? So the Fed is trying to say, "Well, how much should we put on the brakes?" Well, it depends on how much of the inflation is due to excess demand pressures versus supply pressures. And so, that's just a tough task to know in real time. And I think part of the challenge in 2021, why the Fed fell behind the curve, I suspect they were thinking that from a good part of the year that a lot of that inflation was all supply side driven.

Beckworth: They should have, as you just said, been thinking some of it is demand-driven. But in any case, there's just a knowledge problem there. And I think that's a great illustration of why nominal GDP targeting would minimize this confusion. I'm not saying it would solve all the problems, and we have some measurement challenges, how to get better real time measures, maybe forecasts of nominal GDP. But a nominal GDP target, as my listeners will know, all you do is focus on total spending, total aggregate demand and not worry about the month-to-month changes in inflation, which I think we got caught up in, and to some extent we still are. So that's one of my big takeaways, have a monetary policy regime that's robust to big supply size shocks. You're right, we may not have a pandemic again, but we might have wars, something else that comes up.

The bigger question is why weren't professional forecasters worried and why weren't bond markets worried when the standard old fashioned macro textbooks, Keynesians even, if you will, would've predicted a big inflation. I mean, what gives? Why did people not believe in the textbook models? I find that fascinating.

Beckworth: So that's my takeaway. But we'll move on from takeaways to your paper. And I want to talk about your paper and its title is, *25 Years of Excess Unemployment in Advanced Economies.* And it's a really fascinating paper. And Joe, really what motivated this conversation is I heard you, I believe last week at an event at Brookings where you asked the chair about his famous talk where he talked about navigating by the stars, U star, R star, Y star. And in this paper or this talk, he said, "Look, it's hard. It's increasingly hard. You can't really see the star, it's hazy up there, and we need to have some humility about it." So why don't you share with us the question you posed because it's also in this paper, and then maybe as a way to introduce what you were writing.

*25 Years of Excess Unemployment in Advanced Economies.*

Gagnon: Yeah, absolutely. And then we can get to what policies would be better. But let's start with the problem. So, by the way, this paper was written with a former colleague, Madi Sarsenbayev who worked at Peterson for three years and he's now at the International Monetary Fund. But, you mentioned the question I asked Jay Powell last week, and I raised it because if you read the paper you'll see I'm a huge fan of that speech he gave. At the time I wasn't so much, I don't remember what I thought of it. I was cautiously in agreement, but now I'm more strongly in agreement now, because I think he was onto something. I think that central banks all around the advanced economies made a big error for a long time that they didn't see that they had made. And that was what motivated my question because I was worried. I thought he was right but I wondered if that lesson had been forgotten because COVID has upended everything and I didn't want that lesson to be forgotten. So that's why I raised my question. But what is the lesson? Well, all my professional career, I was a macro modeler.

Gagnon: I did precursors to DGE modeling when I came out of graduate school 30 years ago. And so, I could do rational expectations modeling, you remember that? So I was that new synthesis where the Lucas-Sargent met the old Keynesians and that was me. But I remember back then people would sometimes talk about nonlinearities, and I always pooh-poohed it because first of all it makes the modeling so much more difficult if the world is nonlinear, it's hard to solve these things, and you can get any result you want and it puts no discipline, and I didn't like those models. Well, now 30 years later, I think I see a nonlinearity in the world that you really have to think about. And I think COVID actually provides even more evidence for it, but I think the evidence was there even before COVID.

Gagnon: And that is the inflation process, and my view now is about as nonlinear as an economic relationship can be and still be sensible. And that is namely that when you strain capacity, when you reach... There's a certain point at which the effect of extra demand on inflation and the supply curve gets very vertical. And so, any more demand at that point goes a 100% into inflation and zero into output. You just reach some limit of what you can produce, and that's a vertical supply curve there. But then as demand gets weak and so you’re below supply capacity, it gets flatter and flatter. And the reason it's so flat in my view is there's this downward wage and price rigidity. Especially wage rigidity, but even price rigidity, people really, really resist taking wage cuts.

Gagnon: Once inflation gets very low, the distribution of price and wage cuts changes gets very strange looking. There tends to be this big spike around zero with very few negative values, and it really shows how people will quit their job or they'll take unemployment rather than a wage cut. Even though a little bit more inflation and no wage cut will give them the same real outcome, and they don't see it that way. This is just a psychological thing, and I think it's been true for a long time. And if you look at the original Phillips curves, which I think of a Phillips curve as being a labor side implementation of aggregate supply, and what you see is even the original Phillips curve based on the gold pitch standard was highly nonlinear in the way I just described. And I think that we lost that in the period of high inflation in the 70's and 80's, but we've come back to that now. So that is the main point of my paper. And what this does is it makes it possible, because the curve gets so flat when the economy has slack and there's excess unemployment, there's so little downward pressure on inflation in that region that you can hang out there for years. And what happens at central banks, if they see that inflation is stable and it's not too far from target, it's not changing, they conclude that they must be at potential output, they must be at U star or Y star, whatever you want to call it. They must be in equilibrium, so they think they're doing a good job. But what they didn't see was, well, no, that's only because the curve is super flat in this region.

Gagnon: And a flat curve means you can't tell because big movements in output or unemployment have so little effect on inflation that you just don't know. And you see this in estimates of potential output and the natural rate of unemployment. The OECD puts these out, the IMF puts these out, central banks puts these out and they get revised tremendously over time because when you are stuck in a place where there's no change in inflation, then you believe that you must be near potential, and then your view of what potential is changes. And so there's nothing holding that down. So wherever the economy drifts to, the central bank thinks that's okay. Well, it turned out it was not okay. And I think what Powell was onto was exactly this. His idea was maybe we should push unemployment down lower, push output up higher and see where we get to inflationary pressure. And if we don't get to inflationary pressure, maybe we've overestimated the natural rate of unemployment and we should be doing more. I think that's what he said in 2018, and he was right in my view.

Gagnon: And I think this is not just in the US but when Madi and I looked at 11 large advanced economies… and it's starting with this period in which we all went to the low inflation, we all went to a 2% inflation target or lower in the mid 90s. And ever since the mid 90s, this has been a problem. In my view, for 25 years before COVID not a single advanced economy was exceeding its potential output. At best in a few cases for a brief time they might have gotten up to potential, but no one exceeded it for 25 years. So that's just incredible. And if you look at estimates of potential now, they show some asymmetry, some sense in which countries are below potential more often than above potential, but not that much.

Gagnon: But those techniques they use are all biased because they basically don't allow economies to be below potential for more than 10 years in a row. Basically the way they do this, they have to go back to potential at least once a decade, that's just built into their estimating procedures. But in fact, they're forcing something on the data that isn't there. And so we just were below potential for most of the time, and I think this is consistent with undershooting inflation targets for long periods. I think when you get inflation just a little bit below target, people don't revise their estimates of the target. The surveys show that both households and professionals really don't change their estimates of long run inflation very much when you're in the range of target.

Gagnon: And so you can be below target for a long time that people think you're going to get back to target even if you don't. I mean look at Japan, they were at zero for a long time, but expectations were they would be higher, maybe not two, but somewhere between zero and 2%. Year after year, people think that inflation's going to be higher than it is, and that basically enables inflation... You don't get into these deflationary spirals because basically people think it's going to come back and they're willing to be unemployed rather than take the wage cuts that would be needed to get into deflationary spirals. So, the economy can just sit in this position of excess supply and weak inflation but not deflation for a very long time. And economists just don't like that. I mean, if the world is linear that shouldn't happen. If the world is linear, inflation should start falling and falling ever more and you should get these deflationary spirals, but you don't. I think this nonlinearity has important macro implications.

The economy can just sit in this position of excess supply and weak inflation but not deflation for a very long time. And economists just don't like that. I mean, if the world is linear that shouldn't happen. If the world is linear, inflation should start falling and falling ever more and you should get these deflationary spirals, but you don't. I think this nonlinearity has important macro implications.

Beckworth: Yeah, a lot to unpack there, Joe. A lot of interesting observations. Let me begin by going back to this point you made earlier, which is really what I would call the money illusion, that people prefer not to get nominal wage cuts, they'll do other things. And I can speak to this from experience as a former professor, associate professor to be precise, but my experience at universities was that it was easier for the university not to give you a wage increase. Faculty are very sensitive, very sensitive group. So if you were to actually cut salaries, there would be a riot on campus by the faculty. But if you didn't give pay increases and the real wages were falling, they lived with it. So yeah, I can speak to that. That's the money illusion. Now also, I want to go back and just summarize, I think, your key points here, and then maybe unpack them with you a bit more. So you're making the argument that for 25 years, unemployment has been higher than it needed to be to keep inflation low. That's because to some extent we're benchmarking where potential GDP is based on where inflation is, and this is due to Phillips curve type thinking, right?

Beckworth: I'll just throw this out there, maybe we'll come back to it. Maybe this is an indictment of using the Phillips curve. It leads to bad policy outcomes over certain periods. But it is, I know the main macro model people have in their minds. But the first point is, 25 years unemployment higher than was necessary. Second, this arose because we are relying on a linear version of the Phillips curve. So in our minds, we're not thinking things can get nonlinear. I want to come back to that with you. And finally, your solution to all of this, we want to get to the policy part, it's going to be a higher inflation target. That's at least the one way to solve this problem. So, let's go to that first point, and you've touched on it, but let's flush it out a little bit more about unemployment being higher than necessary for 25 years. And you give several reasons, and I think the key one we've touched on already and that is wage and price rigidity, and then also hitting the zero lower bound, creates this money illusion. And it’s money illusion for me the worker, but it's also... If I can use this term loosely, money illusion for policymakers because they're confusing where the real outcome should be based on low, stable inflation.

Beckworth: And this is your concern, but you point to some other deeper things. There's demographics that are causing changes in the stars, right? That's one of your key points. And in fact, in preparing for the show, I went and checked out the FOMC’s… from their summary of economic projections, their long run federal funds rate, the long run unemployment rate, we'd call those U star or R star, and they've all declined dramatically since they started keeping track of this data. So, either those numbers have declined naturally or our understanding of them has failed to keep up with it. But how would you break that down? Has U star or the natural of unemployment, has that fallen over time... I mean, I think you're arguing FOMC officials, they’ve overestimated it, but has it also changed over time? I guess, it would be a related question.

The Movement of the Natural Rate of Unemployment Over Time

Gagnon: Yes, I think both is true and it's not unrelated. I think as it's fallen, the Fed has been behind the curve in not seeing it having fallen. And so that's part of what Powell was saying. But in terms of the people who have done the best work on the natural rate of unemployment are the Congressional Budget Office. And I think the US CBO natural rate is probably the best official estimate of a natural rate I know of for any major economy. And even it's not right in my view, I think it's too high still. But what they say, and I think this is sensible, is that they allow movements in that only for demographics. Their view is that prime age workers want to work, they all want a job or some large fraction of them want a job. And that's pretty constant.

Gagnon: And so that's why a lot of the critics of Phillips curves don't like the fact that, well, why do you assume that this is the supply part of the economy is observable and fixed, and only the demand side moves, right? Because obviously both curves should move. And I think there's a strong case to be made, not perfect, but that... Well, if you control for demographic shifts, then it’s not crazy to think that people's desired labor supply is pretty much fixed. There's a vertical long run Phillips curve maybe or whatever. There's a natural rate of unemployment that is pretty stable once you control for these demographic shifts because prime age workers, especially prime age male workers, they all want a job, and that's just constant. So that's the defense of the Phillips curve.

Gagnon: It's not perfect. I understand. And unemployment isn't the only metric of labor market pressures, and ideally you'd want a better slack measure that took information from participation rates and vacancy rates, but anyway... So a lot of what I say, I talk about the unemployment rate, but it could be recast in terms of some broader slack measure that included these other things. But the point being that there's a sense in which workers want to work almost regardless in the long run. That's what makes the Phillips curve work, at least somewhat. But I grant you it's not perfect. Anyway, the demographics... the biggest thing is the age of the workforce, because as the workforce got younger from the 1950s to 1980, the natural rate of unemployment rose because young people are more likely to be unemployed.

Gagnon: Basically, they don't know what they're good at. They're trying to figure out what their career's going to be and how they can fit in. And the more young people you have on labor force, the more unemployment you're going to have. And that was rising from the 50s to the 80s. And that's partly why we got into the inflationary periods because as the natural rate unemployment was rising, but the macro policy makers didn't realize it, they were trying to hit a target of unemployment that was too low. That's one reason why we got too much inflation. It's not the only reason. Then it peaked around 1980, and then it's been falling ever since. And that's been pulling it back down. And again, if you look at the Fed's forecast and analysis, I don't think they fully saw that process. And so as it was coming down, they didn't see it.

Gagnon: Also, the other demographic changes, longevity and birth rates and stuff, were lowering R star because it was changing the mix of saving and investment. And Fed didn't see that either, and neither did other central banks. And so, these changes, R star and U star, were going on in the background and central banks aren't seeing them, and it's interacting with this very flat Phillips curve segment that they got onto. And they weren't getting a price signal. The inflation rate wasn't showing them the information they needed to see. And since they assumed the Phillips curve was linear and had some kind of slope, they thought they were closer to potential than they were, and they kept making mistakes. That's my view. So it was both demographic shifts... That everyone saw the demographic shifts, but people didn't all immediately then jump to the conclusion of what it would do to R star and U star, but now we’re learning them.

Beckworth: Yeah. So there's been a lot of interesting work on this challenge that the Fed faces and other central banks face in understanding real time slack in the economy, how to measure it, how to get at it. Orphanides has a series of papers where he goes back to the 1970s, and he takes a Taylor rule and he takes a real time measure of the output gap that they had. And he finds, well, actually the Fed was doing a decent job just they didn't have good real time measures. That's his argument, because the argument has been... You know this well, that if you take a Taylor rule, fit it before Volcker, and after Volcker two very different Taylor rules.

Beckworth: But that's using ex-post data after all the data revisions have been made. And Orphanides goes, it's a big deal, and having a good measure of slack, the output gap, unemployment gap. And along those lines, I'm glad you brought up the CBO’s methods because what I like about the CBO, and again, none of this is perfect and there's still a lot of hurdles, but they go from the ground up as you were describing. They try to build up, take the pieces that would make up potential or the natural rate of unemployment from a ground up. Whereas a lot of people, a lot of approaches just take a statistical approach. They just try to fit a line. And that's dangerous too. And to be fair, the CBO doesn't have all the data it probably wants as well to get this perfect measure. But either you can start from the ground up or do a statistical measure, and either way you need some humility in what you're doing. Now in your paper however, you do boldly go where we need to go, and that is to come up with a better measure of unemployment gaps. And can you speak to that exercise and what you found, and how did you do that? How did you come up with a more precise measure?

Constructing a Better Unemployment Gap Measure

Gagnon: Oh, precise is the wrong word to use.

Beckworth: Okay. Okay.

Gagnon: I don't know how to do that but I thought, "You know what? If I write this paper and don't at least take a stab at it, it's... I don't know, it's crying out for someone to take a stab at it." So I did, but it was the last thing I did before publishing the paper and I thought about better ways to do it since, and I definitely think it was only meant to be indicative.

Gagnon: So what we did was we said, "Look, let's make a conservative assumption that countries were at potential at least once in the past 25 years." Maybe they're always below potential, but I think there's some sense in which countries were close to potential at least once. A lot of them, that would've been in 2007 just before the financial crisis, in the US it was in 2000. And in some countries probably it was not at all. Like Japan I don't think was ever, but let's just make that assumption. It's a conservative assumption in the sense that there was never any period after ‘95 where you saw any inflationary pressures. So there was never any period where you could say, we were above potential, but maybe we were close to potential. And those were periods where the problem was inflation was stable throughout, so couldn't really be sure. But that was one assumption.

Gagnon: So maybe we're underestimating the excess unemployment, but at least it's a start. And then the other thing we do is if the unemployment rate at those points is at the natural rate, then that's the natural rate for that country. But we could allow a little slope to the natural rate, a little change over time if the demographics are changing. And basically, we use only the age. And we say, if the effect of the age ratio on the natural rate that seems to be true in US and Canada applies to other countries, then there's a slight slope to the estimated U star from that point. And it goes down over time a little bit, but it's not much.

Gagnon: And so we basically are saying that U Star is a relatively slow moving, very steady variable that only moves a little bit because of demographics, and it's identified by the lowest point of unemployment in this sample. And that's where it touches, and then we say, "Okay, any other movements in unemployment, our reflected gap between wherever unemployment goes beyond those points, raises the gap between that and U star." Then we just drew the graphs to show, and we showed how our estimates compare with the OECD's estimates. By construction basically, they're always above, that's fine. They get close at points, but basically they're always above. And it just gives you a sense of how big this error is. The story's quite different from countries, so I'm happy to talk about different countries, but there isn't one common theme.

Gagnon: The only common theme is that there's simply no evidence that anyone was above potential or that unemployment was ever below the natural rate in any of these countries ever since 1995. There just isn't. And it is hard to know how much the gap really was though, because when the curve is really as flat as it seems to be in this region, it's hard to know. Anyway, I’ll stop there, but I think COVID, for the first time in postwar history, has moved the natural rate of unemployment or whatever you call… the natural slack measure for the economy was moved by COVID in a way that never happened in postwar history. In all of postwar history that was very stable and only detected by demographics. And COVID is the first time something non demographic shifted that U star significantly in many of these countries. It pushed U star [inaudible] unemployment for the first time by a significant amount in a long time.

The natural slack measure for the economy was moved by COVID in a way that never happened in postwar history. In all of postwar history that was very stable and only detected by demographics. And COVID is the first time something non demographic shifted that U star significantly in many of these countries.

Beckworth: Yeah, it's an interesting exercise you do in the paper. And for me what the takeaway is, is we should have humility again. Our estimates of U star are way off... Well, not way off, but they're meaningfully off to the point that they do create policy mistakes. So we need some humility in doing this. And this is a nice segue into a little deeper dig into the Phillips curve itself. We've been talking about the Phillips curve, and it's this relationship between slack in the economy and the inflation. And you alluded to it earlier as a way of thinking about the role maybe labor plays on the supply side of the economy role and how it might generate inflation relative to demand pressures on it. Now, in our principles of macro textbooks, we often talk about a short run Phillips curve, a long run Phillips curve. And I think you also alluded to that when you talked about how you have so much demand pressure, the Phillips curves become vertical. But if you had explained the Phillips curve in a very basic way, how would you describe it to your relative, someone on the street? How would you define the Phillips curve?

Defining the Phillips Curve

Gagnon: To an economist, I would say... And I'll get to a lay person, but to an economist, I would say the Phillips curve is an attempt to get at the supply function, supply curve of the economy by focusing on the market that is the most important by far, and that is relatively stable. There are other features of supply like commodity prices and such that are not very stable. So, the Phillips curve tries to get at the more stable, observable part, but it's really trying to get at the supply curve. That's what the Phillips curve is trying to do. To a lay person I would say, that the Phillips curve is trying to tell us something about the state of the labor market, whether it's overheated or underheated, whether it's adding to inflationary pressures or not. And the idea being that if the economy is just booming too fast and it's going too strong, there won't be enough workers, there'll be people demanding, pushing up prices. I mean, you can't talk about this without using economic terms, so-

Beckworth: Sure, sure.

Gagnon: I've tried, but you get the idea that you could imagine to an overheated economy would cause inflation, basically.

Beckworth: Yeah. And I think there's an intuitive explanation, like you said, it's overheating, underheating, slack in the economy. Now some people, some economists who aren't fans of the Phillips curve, they might say there is some reduced form relationship for sure, but maybe there's a third variable that's explaining both of them, the overheating itself, right? The overheating affects the labor market, also affects the inflation rate. But I think your first point though is there's also a structural supply side curve to the economy as well. So you want to get at that. What's interesting though to me is how you've documented... Others have documented that this flattening of the Phillips curve has occurred for many advanced economies, maybe I should say all advanced economies, right? Since the 1990s. And your explanation for it is wage and price rigidity, is that right?

Gagnon: Yes.

Beckworth: Okay. Combined with policy bringing inflation down. So it was a combination of existing rigidities on the nominal side of the economy, and then on top of that, central banks taking concerted efforts to fight inflation, starting with Volcker. But really in the 1990s you see it coming down and the Bank for International Settlements had a great chart that I saw, it's a few years back but it has that coefficient on the output gap, its effect on inflation. And you see, it flattens. I mean, this is across a bunch of advanced… not just the US. It gets really flat. And it's really surprising. And so, it's interesting to think what has happened over this period. So, let me throw out another potential interpretation. What I'm going to say is actually, I think consistent with your story, but I think maybe it's another way of looking at it. Maybe it's the other side of the coin, I'll put it that way.

Beckworth: But if you go from the 90s, say mid-90s to the present where we have these flattening of the Phillips curve, we have all these other things happening that you alluded to, the demographics, the aging of the planet. And I would also note as a symptom of that, we have the emergence of the safe asset shortage. So, I like to tell this story, Joe, and you can correct me if I'm wrong, but the emergence of the safe asset shortage, I like to frame it in terms of excess demand for liquidity or for safe liquid assets. And if you have this emerge, and I think we do, there's lots of reasons to believe this is a problem and will be a problem after we get to the other side of the pandemic. But this excess demand for liquidity or for money like assets, on one hand it can drive low inflation if it's not being satiated. If we're not providing enough treasuries to the world, not enough moneyed assets to the world, that can lower inflation on one hand. And on the other hand, it can also affect the real economy. It can affect if we're keeping the economy fully employed. So, I like to look at this story. I think it's a complementary story that there's been this safe asset shortage over the same period. And it's maybe just looking at it from a different side. What do you think?

The Safe Asset Shortage Story

Gagnon: I don't have any strong views. I would say, one thing that has always struck me as so odd about the safe asset shortage story is that if that's about US treasuries being the global safe asset, right? I mean, would you buy into that common theme?

Beckworth: I would say it's a big part of it. Yes, the US is providing safe assets to the world and they didn't provide enough.

Gagnon: They didn't provide enough. Alright. And the thing is that... well until recently, it certainly was true in the 2000s and through 2020 anyway. Maybe the last year or two things are changing. But for the past 20 years before COVID, the vast majority of the purchases of US treasuries were by foreign governments or government agencies, it was foreign exchange reserves or sovereign wealth funds and not private individuals. So, the story is often cast in a way that, okay, it's private individuals who want this, economic maximization, but how would you change your story if you found out that... Well, in fact, most of these purchases are not by private individuals subject to market forces, they're by government officials who are not subject to market forces. I mean, that’s a policy decision they made. I think it was because they wanted to maintain large trade surpluses because that was the source of demand that they could engineer for their economies. That's my view, but what do you say to that?

Beckworth: Well, I'd say two things to that. One, so the safe asset shortage does include Treasury securities, but it includes a lot of US dollar denominated assets as well. So, mortgage backed securities, GSEs, even dollar deposits overseas, there's demand for the global dollar system. So yes, Treasury is a big part of that, but it's not the only thing, I'd say number one. Number two though, even when you see that happening, there might be this mercantilist instinct for doing this, but also I think you could see the government acting on behalf of its people. People are trying to save, and maybe it's not the most efficient setup, but the government's effectively doing the savings for the people. And I think your argument is maybe they're forcing excess savings as opposed to more of a natural, endogenous… But I would suspect particularly from Europe and within the US the aging of the population is playing a big part of this.

Beckworth: But I'm not sure it's going to explain everything, but I will note. I do have a paper where I provide some empirical evidence, at least from 2010 to 2019. So I don't go over the whole period where you can view some of this as a safe asset shortage problem. But again, I think, to reconcile the stories, it would be the flip side. Again, I like looking… and this is not everyone in safe asset literature takes this view, but to me, a simple Monetarist story. There's real money demand, it's not being satiated. And by real money demand, I mean everything from currency to treasuries and in between.

Beckworth: So Joe, let's go back to Phillips curves and take a look at some alternative explanations for a flattening of the Phillips curve. So one is that the central banks are doing a great job, they're offsetting shocks in a systematic manner, and it breaks down any relationship between say, the Fed and inflation. If you look at the Fed's interest rate and inflation, you may not see it. And maybe they're also doing in a systematic way where it breaks down the relationship between the unemployment rate. Again, since I threw this money perspective out there, if I think of this in terms of the Fed responding to money demand shocks, even though it's using interest rates, if it perfectly were to offset every money demand shock, you wouldn't see any relationship between stance of policy and inflation, right? So anyways, one explanation is this, there's an identification problem. The Phillips curve is there, it's structurally there. It's just that what we're seeing is reduced form clouded by a really good central bank doing its job.

Alternative Explanations for the Flattening of the Phillips Curve

Gagnon: There’s some papers that say that, and there's a logic to it and what you say is right, but just in terms of in practice, I don't think it's that strong of a critique because what that means in practice... Well, first of all, a lot of that is about... Well, central banks are controlling expectations well, but you can control for that in regressions by using different inflation expectations measures, and therefore that controls for that factor. And then still you should see, is there a Phillips curve, and all you need is, is there any movement in U versus U star or Y versus Y star? And if there is, and you control for the central bank's incredible controlling of inflation expectations, then you should still see the Phillips curve, and you still see it flattening. The other piece is that, well, suppose that we don't have good measures that control for expectations. And that's a plausible argument.

Gagnon: Then what you should see is that the estimates of the slope of the Phillips curve are highly uncertain, right? Those standard errors should get very large. But in fact, we see that the decline in the slope of the Phillips curve comes with reasonably tight standard errors so that we can rule out that it's still steep. It's not like it's a cloud, it's flat, but it's flat within ranges that are still narrow enough. And the story makes sense, but I think the data say… depending on how much you believe the controls for inflation expectations, credibility of central bank policy, but even if we don't control for it well, I mean, the slope is too tightly estimated relative to stories of an old steep Phillip curve.

Beckworth: Alright, so theory is there, but the empirical support is lacking. So another explanation, and maybe this is what you're getting at too, is that paper by Jonathon Hazell, Juan Herreño, Emi Nakamura & Jón Steinsson, *The Slope of the Phillips Curve: Evidence from U.S. States.* And what they argue is that the Phillips curve has been flat, it's not nothing new, it was flat from way back when and they attribute everything to expectations. So I know you've had discussions on this paper. What are your thoughts there?

Gagnon: Well, I have actually been in contact with them, in communication, and I like that paper, but I asked them if they would check for a nonlinear curve. And they said they would look into it. Interestingly, Nakamura and Steinsson have another paper with a different co-author that is all about nonlinear Phillips curves, which I love. And has very different conclusions from the one you cited. So, it's fascinating. But I think I like their estimation, their identification idea of looking across states.

Gagnon: I think if they'd allowed for a nonlinear curve, they might get different results. Instead of a linear curve with a modest slope, they'd probably get a nonlinear curve that had a steeper segment and an even flatter segment, but they haven't tested that. But their other paper assumes exactly that, that the curve is extremely nonlinear. And then they show what that means in macro things… and that actually is the most important theoretical underpinning of my paper, Madi and I… I'm not a theorist and they are, but they show in theory how this all works, and so I rely on them. So it's interesting you raise them because they have this two papers.

Beckworth: They have another paper. Yes.

Gagnon: The two papers are in conflict and they themselves need to decide what they believe.

Beckworth: So, which paper they believe in? They have one paper that's linear, one that's nonlinear. And Joe, the world is nonlinear in many ways, right? I mean, when I explained to my son why he should take calculus, I'm like, "Look, the world's not linear, you got to wrestle with curves in the world, nonlinear relationships. That's why we have calculus.” Yeah, so that's a good point. And I think as economists need to wrestle with nonlinearities too, better. And I know back in grad school, I remember my time series class, being exposed to these threshold models, these regime changing models. Things do change. So let me throw out another objection. This isn't to flat Phillips curves, but one objection I get to this measure which I show, which is... By the way, the nominal GDP gap is based on consensus forecasts. Their forecasts of nominal GDP are usually above what actually transpires. So they're persistently overestimating how strong the economy's going to be.

Gagnon: How long is that true?

Beckworth: All the way back to the end of the great financial crisis, so the past decade. So the point is, you have a whole decade where they get their forecasts wrong, and some people have a problem. Are you telling me they systematically get things wrong, or alternatively I'll get to asked this question, what about the classical dichotomy? Are you telling me the Fed can keep, for decades, the real economy artificially low? Have you had that critique asked about your view here that the Fed is keeping unemployment below for such a long period of time, does it really have that control?

Gagnon: I haven't heard that, but I guess it's the whole point of the paper is precisely that that argument is wrong. That that argument is true in a linear world, or at least a linear world in which there's some slope to the Phillips curve, but in a nonlinear world in which part of the Phillips curve is extremely flat, then it's not true. Now, I think a different critique in my paper would be, we just think the slope of the Phillips curve shifts from time to time, and it's very flat now, or it was very flat in your period. And we don't believe it’s non-linear, but I don't think that even makes any sense. And when you ask these people, "Well, what do you think would happen if demand was 20% higher than supply? If nominal GDP jumped 20%, do you think that the economy would just produce 20% more or do you think prices would start to rise?”

Beckworth: Right. Right.

Gagnon: Then they always say, "Oh well, there's going to be some point at which..." Well, that point is where the nonlinearity kicks in.

Beckworth: Very intuitive. Yes.

Gagnon: I've actually presented this a couple times and I've actually talked to people about it. And what really has surprised me more than anything is that almost everyone nods their head and says, "Yes, I think that's right." And I'm like, "What? You believe me?" I’m so used to people objecting and it's really like, I’m really surprised at how little people have pushed back. They just say, "Yeah, doesn't everyone know that?”

Beckworth: Isn't that obvious? Yeah. Well, let's move to the policy implications in time we have left. So walk us through what this means for the Fed moving forward.

Policy Implications for the Fed Moving Forward

Gagnon: Okay. So I have in the paper two points, but I'd like to also talk about the third, which is how would nominal GDP targeting improve things?

Beckworth: Okay.

Gagnon: Because I think nominal GDP targeting would help a bit, but in my view it would need to come with a higher growth path to be as effective as you would like. But I think nominal GDP targeting would help but it isn't enough or it would take too long to really make a difference. So, the two points we make in the paper are that central banks should have more humility about their models. This is an important non-linearity, which they often don't have in their models. They should keep this in mind when they try to estimate the natural rate, when they try to think about where they are. And even with or without that addition to their model toolkit, they should have some more humility about where the natural rate is. And if the mandate is maximum employment, maybe they need to experiment every once in a while to see how high they can push employment, how low they can push unemployment without causing any problems to the other side of the dual mandate.

They should have some more humility about where the natural rate is. And if the mandate is maximum employment, maybe they need to experiment every once in a while to see how high they can push employment, how low they can push unemployment without causing any problems to the other side of the dual mandate.

Gagnon: That's a strange thing to say in a purely linear, moderately steep Phillips curve world because it's such a trade-off between the two. But I think in this world, you have at least a region of the world in which there's almost no trade-off. And I think that's key. But the other point is that all of this would be easier if we had a bit more inflation, because this nonlinearity and extreme flat part of the curve is created by having a very low inflation target. And if you had a somewhat higher inflation target, you could get away from that a bit. It would be easier to know where the natural rate of unemployment is. It would ease the adjustment in labor market. And this is the Nakamura & Steinsson point, which I think is true but I have less evidence for. I think it's possible that a too low inflation target raises U star itself directly, because it means that you have to have even a bit more unemployment to keep people satisfied with such low wage increases.

Gagnon: In other words, because of money illusion, you are going to need a bit more unemployment to keep people at the very low rate of pay increases that you need to meet your inflation target. And if you have a higher inflation target, then you can get away with less unemployment because the people who are getting pay cuts won't resist so much. That's the Nakamura & Steinsson point of their second paper. I think there's some evidence for that, it's hard to really show. Akerlof, Dickens and Perry had a paper that said the same thing 20 years ago. They didn't think it would really kick in until the inflation target got below 2%. And so, we're right in that region where it could kick in. What I see more evidence of is the first point, which is that central banks could have lowered unemployment even with the inflation target they had, but there's a possibility that they could lower it even more with a higher inflation target. So those are two things.

Gagnon: Well what about nominal GDP? I keep expecting David, you, to ask me, wouldn't a nominal GDP level of target eventually correct this problem? Because if there's at least some slope to the Phillips curve so that... If U is above U star or Y is below Y star, inflation will undershoot the target a bit, and therefore you're going to fall below your nominal GDP path. And then, eventually that gap's going to widen enough that you're going to have to do some policy to accelerate growth, and that's going to get your unemployment rate down as you accelerate back up towards your NGDP path. And you'll solve that problem eventually because you'll see that you're drifting farther and farther away from the path. And I think that is right, and that's why I would support an N GDP target. But I think how right it is and how effective it can be depends actually on the slope of the NGDP path.

Gagnon: Because say, if you have a path like 4% which is consistent with 2% inflation, the errors that you see are going to be very small at first and creep in only slowly. And so, it's going to take longer for it to kick in. A higher path, if you had a 3% inflation target and therefore a 5% NGDP path, I think you'd be seeing deviations sooner. In other words if unemployment was above U star, in that world, you'd be on a less flat part of the Phillips curve because you'd be away from the Zero Bound and everything, and therefore you'd start getting more disinflation and you'd deviate from your path faster. And so the policy reaction to get you back to the NGDP path would kick in sooner, and would be more robust. But I think either way, I was convinced you were going to tell me this or ask me… this was going to be your point: “Well, wouldn't an NGDP path solve this problem?” And the answer is "Yes, David, it would solve this problem." But it will solve the problem better, more reliably, more quickly, if the path is a little bit steeper than the 2% inflation path would imply.

Beckworth: Joe, I will gladly take a 5% nominal GDP growth path target. So, happy to sign you up. One last question related to this, and that is we've gone through this high inflation period. And again, as we mentioned earlier, some of this is due to supply side shocks, and all of it is due to too much aggregate demand. But with that said, from Congress’s perspective, from the body politic’s perspective, we've just had a period of high inflation, some policies somewhere screwed up and we can't let this happen again. And my concern is that it may hamper or hinder future attempts at makeup policy. So key to what you've just described there, that nominal GDP target is a level target where you make up, right? And that's my concern is that we may have cold feet moving forward. The Fed itself may have cold feet, even with FAIT. Will the Fed really try to run things hot again, given all the heat they received? No pun intended there. That's my concern there is hopefully this experience doesn't undermine... I mean, decades of research on level targeting, makeup policy, the baby doesn't get thrown out with the bath water because of this one bad experience.

Gagnon: Yeah. Well, I hope so too. I think that's exactly what I worry about, and I keep thinking COVID's going to go away, and I thought so a year ago and I was wrong. And so, here's another confessing up to an error. But I continue to believe that COVID is not a permanent change in how the world works, and it's not likely to happen again for 30, 40, 50 years. So, I hate to have it change how we think about policy too much, because I just have this feeling that it's a shock, it's not a structural change, but maybe it is. What do you think do? Do you it's a structural change?

Beckworth: I think it's a shock, it's just a long-lasting shock. I don't think it's going to fundamentally change the key drivers of the economy. I think those are already set in stone, the demographics, things like that. Well on that note, our time is up. Our guest today has been Joe Gagnon. And Joe, thank you so much for coming on the show again.

Gagnon: Oh, you're welcome.

Support the podcast by making a donation during this holiday season!

Photo by Chip Somodevilla via Getty Images

About Macro Musings

Hosted by Senior Research Fellow David Beckworth, the Macro Musings podcast pulls back the curtain on the important macroeconomic issues of the past, present, and future.