John Roberts on Macroeconomic Modeling at the Fed, Makeup Policy, and the Future of FAIT

John Roberts is a 36-year veteran of the Federal Reserve Board and mostly recently was the Deputy Associate Director in the Division of Research and Statistics, overseeing the board’s domestic macroeconomic modeling efforts. From 2017-2019, John also served as a special advisor to Federal Reserve Governor Lael Brainard, where his responsibilities includes preparation of speeches, providing advice on monetary policy, macroeconomic forecasting, and regulatory attending FOMC meetings. John joins Macro Musings to talk about his time at the Fed, macroeconomic modeling at the institution, his work on the zero lower bound, and current Fed policy. Specifically, David and John also discuss the art of interpreting the Fed’s Summary of Economic Projections, the future of modeling for policymakers at the Fed, the state of FAIT at the central bank, and a lot 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: John, welcome to the show.

John Roberts: Thanks, David. Looking forward to it.

Beckworth: Glad to have you on. And I have been following your blog, John Roberts Macroeconomics, so I would encourage listeners to check that out as well. We'll have the link to it in the show notes, and we'll come back and talk about it later on. I have a few questions about your blogging. You've stayed busy since retirement, but I'm curious to hear about your time at the Fed, 36-year veteran. That's quite impressive. So walk us through your journey at the Fed.

John’s Professional Journey at the Fed

Roberts: So my very first day I stepped into the Fed. I was 22 years old right out of college. So I had a job where there are lots of opportunities at the Fed as a research assistant. And going in, I had been a math major, so I already had this interest in analysis. And then I liked the idea of public service. It actually worked just like I was hoping. So the Fed is really this terrific combination of analysis and opportunity for public service. And then in addition, there are lots of colleagues with whom you work closely. That's not always the case for a lot of economists where in academic departments, I think, people are a little bit more isolated, so that's maybe not the right word. But anyway, very collegial at the Federal Reserve. So that gives lots of opportunity to work with other people, form friendships. And then as part of that, as my career advanced, at one point, I was a manager with 20 people under me. And that was like a really wonderful experience because you have this opportunity to mentor people, to help them develop. And I found that a really rewarding thing.

Roberts: As you mentioned, a lot of my career was around using models to answer questions of interest to the members of the Board of Governors and to the FOMC. So for example, we would regularly present to the policymakers alternative scenarios, so people would come up with a judgmental forecast of what was going to happen to the economy. And then my group would use models to explore alternatives around that baseline forecast. And then we would also use models to take systematic approaches to monetary policy. So what that would mean is that we would treat the Federal Reserve like it was one of these textbook optimizing agents that economists like to talk about so much.

Roberts: So we would say that the Fed has these objectives given to it by Congress of maximizing employment and achieving price stability. And so we would treat that as a rigorous optimization problem and then give policy advice based on that. Of course, the policymakers are free to do what they want, but this would be an input into their policymaking process. And as you mentioned, I worked for a couple of years for, then Governor, Brainard. She's now the vice-chair of the Fed, and that was a terrific experience. I got to see the inner workings of how policy was made, really a fascinating experience.

Beckworth: That's great, 36 years at the Fed. So you came right out of college, you said, with your undergrad, so you must have gone back and got your PhD then and continued working.

Roberts: Yeah, I did. I worked for a couple of years. When we have people start as research assistants, our dream is always, of course, that they'll, after a couple of years, go away, get a PhD, come back.

Beckworth: Does that happen often that you get RAs that come back?

Roberts: It happens often enough, and people can go on to have very good careers at the Fed. But people do other things as well, right?

Beckworth: Yeah.

Roberts: They typically will go to graduate school of some kind after the Fed, often to get an MBA or a public policy degree or something like that, or people will just find other jobs. But certainly, it's a great opportunity for undergraduates who are interested in economics or analytical stuff.

Beckworth: Sure. So a couple of observations, so you mentioned the neat environment of having lots of other macroeconomists around you. I can speak to that. I worked at a state university. There's only a few other macroeconomists in the department, so you're right. We look forward to conferences where we get together, big gathering of fellow macroeconomists. We can correspond via email. Twitter's been great when I was an academic because I was able to interact more on social media. But I think that environment would be amazing to be in a world where you got lots of macroeconomists to bounce ideas off. And I noticed in your resume, you published in some top journals, the AER, JMB, JMCB. So you're working with your colleagues, co-authoring papers. It sounds like a really exciting environment. I’ll also mention all these scenarios that you did, this macro modeling. It's great for someone like me to go back and read the transcripts, and we get to see some of these efforts. So we see these different scenarios, different paths, and I imagine you're the author behind those.

Roberts: I've played various roles over the years from being the person running the software to managing the people doing that to managing their manager, but yeah, very much involved in that process at various times.

Beckworth: And then working for a governor, Lael Brainard, did you have an office near her? How did that work? Would you visit with her every day? What was the arrangement?

Roberts: Right. So the governors are on the second floor, and a lot of the staff is on the third floor. So I would just be running up and down the stairs frequently, so I didn't have to change offices. It was cool. There's a security for the second floor where the governors are, but I had a special card that would allow me to get in.

Beckworth: Nice.

Roberts: Yeah.

Beckworth: As they say, access is power.

Roberts: Oh, absolutely. Absolutely.

Beckworth: Well, I had Randy Quarles on the show, and he told an interesting story that he was working late because apparently, he commuted, and he still had his family back in Salt Lake City. So he didn't really have a home to go to in the evenings. He stayed in a hotel or an apartment. But in any event, so he'd work late in the evenings, and he didn't realize that there was a panic button underneath the desk, and so his knee would hit it. And the first time he did it, the SWAT team comes running into his office, and he ended up doing it multiple times. He kept forgetting and he'd hit it but interesting story about security on that second floor. They really look after their governors.

Roberts: It's too bad that the world is like that but that's...

Beckworth: Yeah, that's where we are. So you went to the FOMC meetings with Governor Brainard. You talked shop with her. You would write speeches. So how did that process work? Would she come to you and say, "Hey, I want to write on X, Y, Z. Give me a rough draft. Give it to me." How did that iterative process work?

Roberts: So I think each of them is different. The approach with her is that she'd like to... There would be early preliminary stuff with an outline, and she'd give input on that. But then she would pretty much want someone like me or someone I would be working with to give her a complete draft, and then maybe take her edits and do a second draft. But then at the end, she would make it her own. So in terms of ideas, she was pretty open to what I and other staff members had in mind. But there was definitely a process of making the speech her own. She was herself a quite very good writer, so that worked well. It was great to see. Often, I could see my ideas would definitely be there in the speech, but of course, with her words.

Beckworth: Absolutely. So John, you've had a great career at the Fed, again, 36 years. Any career advice you would give to a young and budding economist wanting to work at the Fed?

Roberts: So obviously, if you're studying money or macro, those are the obvious things that folks are looking for at the Fed. But in fact, they hire folks from a really broad range of subject areas. Don't rule yourself out if you do labor economics or even environmental economics. It's a big place and there's lots of opportunities. The process is pretty simple. Young economists, people who are in PhD programs are familiar with how the job market for economists works so just apply and then take it from there. If you're really interested in working at the Fed, well then pursue summer opportunities or there's scholar positions for people in PhD programs because that obviously gives a strong signal of your interest.

Beckworth: Sure. So follow the John Roberts model, become an RA.

Roberts: Yeah, absolutely.

Beckworth: Get some experience and then you'll know for sure whether you want to work there and then come back after your PhD.

Roberts: Absolutely.

Beckworth: Well, let's move on from your career at the Fed to what you're doing now. I want to go back to your blog. It's John Roberts Macroeconomics, and the actual address is jrobertsmacroecon.wordpress, and we'll have a link in the show notes so you can find it. But you are staying active there, and I've been reading it. And what's neat is every time there's an FOMC meeting, John, you put up an interpretation, at least, the ones where they have the Summary of Economic Projections. You put up an interpretation, and I think you're using the standard New Keynesian model to do this. So you help us understand, what are they thinking? What do these changes mean? So tell us about your work there on the blog.

Diving into John’s Blog Work

Roberts: When I was first starting out blogging a couple of years ago, I wanted to do something that nobody else was doing because it's more interesting that way. So it seemed like there was this opportunity, that the Summary of Economic Projections comes out four times a year. And an interesting question is, well, does it hang together? So there is this issue that the summary of economic projections is, at this point, 19 different projections. And what people focus on is the median of those 19 projections. So that need not be a coherent forecast because the median on the unemployment rate might come from a different person than the median on the inflation rate, for example. So that already would raise red flags for some people. But I've forged ahead, and I've been finding that it's actually pretty easy to use a standard model that I'll talk about in a minute to interpret the median from the Summary of Economic Projections.

Roberts: So what does that involve? So as you suggest, David, I'm starting with a pretty standard model. So for macroeconomists, you'll know this as the three-equation New Keynesian model. And so I'm starting from that, and then I'm using a long-term interest rate in my main equation for output rather than a short-term interest rate. But you can see if you click the links on my blog how those things are related. I think the long-term interest rate is interesting because the consensus is that that's what consumer spending or housing investment mostly responds to.

Beckworth: So this is in the expectational IS equation?

Roberts: Exactly. And again, if someone actually is a PhD economist, they can easily understand that relationship. So I have this model that is, as you say, very close to the model that a lot of economists are very familiar with. And then the interpretation comes in… okay, so what parameters would you need to make the thing hang together? And then again, going into some jargon, what sort of shocks would you need to make things hang together? And so what I try to do is to pick a small a number of shocks as possible, try to be economical in the number of degrees of freedom I give myself to try to explain the forecast, then tell the resulting story.

Roberts: So for example, one thing that's really important in the SEP, the inflation forecast is always coming back down to 2% in the long run. And for a long time, it was doing that with the unemployment rate just gradually going to the Fed's long-run estimate of where it was going to be. So how does that all add up? What I interpret that as meaning is that the FOMC participants are assuming that the public's long-run inflation expectations, where the public thinks inflation is going, is the same as their 2% target, that that seemed to make the forecast hang together. You had to make that assumption. And then once I made that assumption, then I could notice that inflation was gradually coming down, but it wasn't coming down right away. So that was suggesting that, in terms of the model, there was a fair amount of persistence in inflation, but not so much persistence that those long-run inflation expectations were moving around. So that gives you a flavor for the kind of things that I do. And if folks want more detail, it'd be great if they check out my blog.

Beckworth: Yeah, check it out. It is interesting because sometimes you do wonder what exactly is the Fed saying or signaling by this SEP? And to be clear, this is a conditional forecast. It can change, right, going forward?

Roberts: Oh, absolutely.

Beckworth: It's a median. So you got to be careful there too as you alluded. But sometimes people look at this and they wonder, and I know that the reporters, they asked Chair Powell at the press conference, “Can you help us understand what this means?” Let's do an example like the December FOMC. It shows unemployment going up quite a bit, at least a percent I believe, or so.

Roberts: Yeah.

Beckworth: And then inflation still slowly coming down. So is the Fed forecasting a recession? How do I understand that increase on unemployment as the inflation rate slowly comes down?

Interpreting Macro Indicators in the Summary of Economic Projections

Roberts: So when they first put that feature in, which was back in September, I had to make some adjustments to the way I was thinking about things. So it had been the case that, as I was saying a second ago, that the unemployment rate would gradually go to 4% which is where they publish where they think things are going in the long run. And for the unemployment rate, that was 4%. And so it was gradually coming up from below. But then in September, for the first time, they had it overshooting the 4% going up to close to 4.5%.

Roberts: So at that point I decided that what they were assuming was that temporarily, what economists like to call the natural rate of unemployment, was temporarily higher than that long-run value. So I inferred that for a couple of reasons. First, it made the story easy to tell, but it also lined up with what some of the members of the FOMC were saying in their speeches. So at a couple of press conferences, for example, Chair Powell had said that he personally thought that, for various reasons, that natural rate of unemployment was temporarily above 4%. And this gets to this whole story around how tight the labor market is right now. Wage growth has been quite high.

Roberts: So we can go back right to pre-COVID. Pre-COVID, the unemployment rate got down to around 3.5% where it is now, but inflation was quite low, and wage gains were pretty modest. They were around 3%. Fast forward to the past year or so, when inflation has been at three 3.5%, the labor market has looked a lot tighter recently at that same unemployment rate than it had looked pre-COVID. COVID messed up a whole bunch of things in the economy. In addition to the high-wage growth which Chair Powell has pointed to as one reason why you might think the labor market is tight, you can also look at some other indicators of the labor market like people quitting their jobs and finding new ones right away. That's an indication of a tight labor market or these job openings, that firms are having trouble hiring people and so they're posting lots of jobs. That's another indication of a tight labor market.

Roberts: So it seems like because of all the transformations in the economy that have gone on, probably because of COVID, it looks like the labor market at a 3.5% unemployment rate is a lot tighter now than it was pre-COVID. So that's consistent with this story that Chair Powell was telling that temporarily, this natural rate of unemployment, the unemployment rate you need to stabilize inflation, is temporarily high. And then there's other work that's come out of the Fed as well. Governor Chris Waller worked with one of my former colleagues, Andrew Figura, on some analysis trying to explicitly link this high-vacancy rate to what that might imply for the natural rate of unemployment. And I've talked about that a little bit on my blog as well. So 4.5% is a reasonable guess, maybe a little on the low side, but a reasonable guess of what the natural rate of unemployment might be right now. And so in the December SEP and starting in the September one, the unemployment rate is going up to 4.5% and then inflation is gradually coming down. So the story I started to tell at that point was that the natural rate of unemployment was higher, and importantly, that that's probably what the FOMC median person was thinking.

Roberts: An important aspect, I think, to keep in mind about the SEP, and one of the things that makes it so valuable is that you were saying, David, that it's a conditional forecast. But one of the things it's conditional on is their own view of what the best monetary policy should be. So when they're setting that path for the federal funds rate in their economic projections, it's their own view of what they need to do to accomplish their goals of maximum employment and price stability. So when they decided that the federal funds rate needed to be high enough to get that unemployment rate up to 4.5%, when they were writing that down, it was because they thought that's what they needed to do. So that's one of the reasons why I invest so much time in analyzing the SEP because I think it's a really good window into what it is that the FOMC is thinking.

Beckworth: That's a good reminder. It's a conditional forecast conditioned on their view of the appropriate stance to monetary policy.

Roberts: Yes. And looking out the window, seeing where things are today.

Beckworth: Yeah, currently.

Roberts: Exactly.

Beckworth: And that makes sense to your explanation that their view of the natural rate of unemployment has changed, at least, temporarily. So 3.5% today is not the same thing as 3.5% unemployment in early 2020. So things have changed dramatically. But just to be concrete and specific here, if in fact that does play out, we go from 3.5% to 4.5% unemployment. Is that going to be a fairly painless process? Or does that mean we're going to lose a lot of jobs? How does this unfold do you think?

Roberts: It's absolutely unfortunate for people… that 1% point on the unemployment rate means that millions of people lose jobs, so clearly not a good thing. But remember as Chair Powell keeps emphasizing, it's really important to achieve price stability, if the longer that high inflation goes on, the possibility that those inflation expectations could get ingrained into people's habits, if that happens, then the pain, the high unemployment that you would need to achieve price stability later on, would be even worse. It's better to have maybe a small recession now than a really big one later on. So yes, it's not good that the unemployment rate is going up, but price stability is, first of all, part of the Fed's mandate as well as maximum employment. And even in terms of the maximum employment goal, the costs are probably going to be lower today than they would be if the Fed did didn't do anything.

Beckworth: So in macro, one of the key things we always need to keep in mind is to do the right counterfactual. It's always easy to say, "Hey, why not do this or this?" But in the case of what you've just said, it is easy for critics to say, "You just want to create unemployment, millions of jobs." But the counterfactual is not a million jobs versus nothing. It's a million jobs versus many more in the future. The right counterfactual is that we have this tightening now versus even more severe tightening in the future. So we got to keep the right counterfactual in our mind. We'd rather do it now than in the future. Well, let's transition from that to the model we were just talking about. I want to spend a little more time on this New Keynesian model, the three-equation standard workhorse model. If you've gone to any kind of graduate program and you've taken macro courses, you've seen it. And for those who've only had undergraduate, you can think of it as a model with an aggregate-demand equation, aggregate-supply equation, and a monetary policy equation. Aggregate demand would be the expectational IS equation.

Beckworth: The aggregate supply would be the Phillips curve. And then the monetary policy equation would be the Taylor rule. And then there are other things lurking in the background that you assume away such as money demand. It's relegated out of the pitcher. But those are the three equations. They're the workhorse. And so here's a question that I often wonder. So people like you who are smart, well-educated, who play in these models all the time, you think through them. You can probably dream about these models, John. But does Chair Powell, when he sees the newest data, he sees the employment data, he sees the employment cost index data, he sees forecasts. Does he think through these models in his mind? Or is he thinking through this as maybe a simple Phillips curve relationship?

Economic Modeling at the Fed

Roberts: So let me start with a couple of other either current or recent FOMC members where I can state with absolute confidence that they can and often do think in these terms. So for example, Richard Clarida, who had recently been vice-chair of the Fed, Charlie Evans, who is just now leaving as the president of the Chicago Fed, they were actually instrumental in the development of that standard model that you were talking about. So they, for sure, can think in those terms. Others who come to mind right away would be, for example, John Williams who had been my colleague for a long time on the Fed staff; Jim Bullard, the current president of the St. Louis Fed; Chris Waller, who's on the Board of Governors. These folks are also professional economists who would be very familiar with these models.

Roberts: So these folks could easily think about policy through that lens, but none of them would be completely wedded to it. And they could also think about policy in other ways as well, and through the lenses of models that might be more elaborate than that really simple one. So Chair Powell… So Chair Powell, he's not a professional economist, but he's also really smart. So he is, I'm sure, conversant in these kinds of models. He has a really big staff that he consults with all the time. So while he's not like a John Williams or a Richard Clarida who eat and breathe these models, he's a smart policymaker who knows how to think through the issues and talks a lot to people like Rich and John.

Beckworth: Well, let's talk about another model that has emerged over the past few years, and that's the HANK model. So the HANK model, it's an acronym, it's Heterogeneous Agent New Keynesian model. And the one we were just talking about has been called the RANK model, the Representative Agent New Keynesian. So in the standard New Keynesian model, we have this representative household. So it stands in for every everybody else. But the HANK model says, "Hey, let's take seriously the notion that households are different." And if we do that, you get some interesting results. You see, for example, people who live paycheck to paycheck. They can be poor people. They even do wealthy people who live paycheck to paycheck.

Beckworth: And when you bring this in, you get things such as, intertemporal thinking isn't as important for people living paycheck to paycheck. They're not looking at the interest rate and making a savings versus consumption decision like you would with the Representative Agent New Keynesian model. But nonetheless, it's a model that's come in. It's popular. I've had a guest, Ben Moll, on the show who is a big proponent of it, he’s at the London School of Economics. But my question is, how relevant are these models to central bankers? Do central bankers use heterogeneous models at all and in particular HANK models? Where are they when we think about these policymakers making decisions?

Roberts: So the HANK models themselves can have some technical difficulties involved with them because instead of the representative agent model, you've got one person making the decisions in the economy. In the HANK models, you've got lots of different people making decisions. It makes it technically more difficult to solve the model. So that gets a little bit in the way of they’re being used in a widespread way. Now, just the heterogeneous agent aspect of it is very prominent in central bank analysis. There's a current hot topic is this issue of all of the improvement in the balance sheets of lots of different folks in the economy during the COVID crisis. There were a couple of different reasons that people ended up with a lot more money in their checking accounts than they had before. One of them was that during the lockdown, people couldn't spend money. The people who still had their jobs couldn't spend their money on anything. And then as well, there was a lot of fiscal largesse that people also couldn't spend because the economy was locked down. So peoples’ balance sheets were in pretty good shape.

Roberts: And so what folks have done is to look at the balance sheets of different groups of people to try to get a sense of, how long can people keep spending out of their balance sheets? And that's been important over the past year. As monetary policy has been tighter, consumer spending has held up remarkably well. And people have been looking at these individual households and saying, "Ah, yes, but we can see why that's the case because the balance sheets of individual groups are holding up well." So let me pivot back to the HANK models. So it turns out that early on in the COVID economy when some of these fiscal policies were first being put into place, some of these HANK modelers said, "Aha, I've got just the model for this." And I was talking to one HANK expert, a guy I used to work with at the Fed, Chris Carroll, who's now a professor at Johns Hopkins. And he pointed me to this model that he put out in 2020, right in the thick of things, that actually predicted that people would not spend everything that they got from the fiscal programs, that they would save some of it, even folks who you might think of as being paycheck to paycheck, even some of those folks saved a substantial chunk of it. And so he actually predicted a lot of what subsequently happened. So that was kind of a win for HANK models, I think.

Beckworth: Yes.

Roberts: So I think, for sure, HANK models are part of the discussion, but notice that that was a fiscal issue rather than a monetary issue. So I think they've really shown their value, I think, for analyzing fiscal policy where issues around distribution are really important. I think for day-to-day analysis of monetary policy, my understanding is that folks are still more in a representative agent-type world. But as computers get faster, it's possible that eventually all models would be HANK.

Beckworth: So I think probably one of the big constraints is tractability, being able to process a bunch of new information in a model. And so it's easier to maybe think in terms of simpler models. But with that said, I want to throw something out there, maybe a shot across the bow of the HANK models here. This is a paper from Eric Sims, he's a past guest of the show. He's a professor of macroeconomics at Notre Dame, writes for the NBER. He’s had lots of great work he's put out with two of his colleagues. And I just mentioned Eric because he's been on the show. But the title of the paper is *Unconventional Monetary Policy According to HANK.*

Beckworth: Now, let me just read the abstract which is interesting. It says, "This paper studies the implications of household heterogeneity for the effectiveness of quantitative easing or QE. We consider Heterogeneous Agent New Keynesian model, HANK, with uninsurable household income risk." I'll skip some of the abstract and jump to this point. "We find that macro aggregates react very similar to a QE shock and a HANK model compared to a Representative Agent New Keynesian model or RANK model. This finding is robust across different micro or macro distributions of wealth." So one of the implications is even if you go to all the trouble, sometimes you get the same result. Now, it's good to know the details. Maybe there's political economy questions you want to consider and stuff. But I'll throw that out there. So again, there's tradeoffs between tractability, usefulness when you're in the policymaking world. So we've talked about RANK, talked about HANK. Anything else we should be thinking about? What is the future of modeling for policymakers? Are they looking at other models?

The Future of Modeling for Policymakers

Roberts: So the model that's most used for the domestic policy work is a model that we affectionately call FRB/US which stands for, the FRB part is Federal Reserve Board, and the us is US, the FRB/US model. And it's a relative of these New Keynesian models. It's a very big model. It shares with the key elements of the New Keynesian framework, are that people are solving optimization problems, textbook optimization problems, and consistent with that, they have model consistent expectations or what's sometimes called rational expectations, so a very rigorous approach.

Roberts: And the FRB/US model has optimization problems. It has forward-looking agents. But sometimes to make the data fit, it adds other elements as well. So a model that's adjacent to New Keynesian models, but bigger and maybe more realistic. So that's the model that, when I was talking at the beginning about what we often use for these alternative scenarios or the policy analysis, that's the model that's been used for a very long time. More recently, it's been supplemented with models that are closer to the New Keynesian ideal where everybody's optimizing, and there's purely rational expectations. So both of those kinds of models are in use at the Fed.

Roberts: Going forward, I think that a fertile ground for further research would be adding more behavioral aspects to our models. To non-economists, it's perfectly obvious that people aren't perfectly rational. Economists are still coming to grips with that. But I think all these financial crises and things like that that we've had over the years should start to convince people that these perfect rationality models probably aren't the most realistic. So incorporating aspects of departures from perfect rationality are probably a way to go in economic modeling. I've done some work in that direction myself and others have as well. I think that's fruitful.

Beckworth: Yeah, we'll talk about that in a minute and your paper with Michael Kiley. You guys do that. It was neat. But I just want to go back to the FRB/US model.

Roberts: Sure.

Beckworth: And it's really interesting to hear you say that. I think it's the first time I've heard this, because you often hear, quite honestly, derogatory comments about the FRB/US model. You hear something along the lines, "Oh, that's something Lucas discounted back in the '70s. It's just a large system of equation models." But what you're saying, it's actually not. It has rational expectations, at least, in part of it. It's not just the old econometric macro models from the 1970s, something different than that.

Roberts: It was developed in the '90s and it knew about the Lucas Critique when it was being designed. Some people, some prominent macroeconomists will acknowledge that the FRB/US model is similar to a New Keynesian model. But others have criticized it as being too close to those models of the '70s. So that's okay. Being criticized by both sides is often a good place to be.

Beckworth: A good sign.

Roberts: Right.

Beckworth: It's a good sign. No, that's good to know. So it is something useful. And again, you want to have maybe multiple models when you're thinking about a question.

Roberts: Absolutely.

Beckworth: It's neat to see this. And I've seen some of your colleagues who've left the Fed, they will use it in a simulation, they'll ask a question. I think I've seen a David Wilcox paper. He's over at, I believe, the Peterson Institute. He did a paper thinking about the average inflation targeting framework. I believe I saw Andy Levin use the FRB/US model for simulation in a paper. So it's still useful and it's good to know. I also mentioned to our listeners, you can actually download this if you have EViews, I think. You can download and run it on your own. Is that right?

Roberts: We've recently posted a version that works on freeware.

Beckworth: Oh, okay. Nice.

Roberts: So you no longer need to buy a commercial product to make it work.

Beckworth: Oh, very nice. Okay, so let's move from models onto your work. Actually, we'll use this segue of models because this work that you did with Michael Kiley, one paper with Michael Kiley at the Brookings [Institution], *Monetary Policy in a Low Interest Rate World.* Then also one of Michael Kiley and Ben Bernanke, similar title, *Monetary Policy Strategies in a Low Interest Rate Environment.* At least in one of these papers, I know you do this. You take an approach for it. The model has consistent expectations across everybody, households and markets. Then you have one where only the markets are really forward-looking in a consistent… and households are more maybe backward-looking or adaptive expectations. So you put what you just said to work. You actually do it. And I like this because the lower-bound issue has been such a big issue and the decade leading up to... Even before that, I guess, late 1990s in Japan, but definitely the last decade leading up to the pandemic for the US. And maybe we'll come back and talk about whether we think we're going to return to that world in a bit. But tell us about this work. What did you guys do in these papers? And what did you find?

Monetary Policy in a Low Interest Rate Environment

Roberts: So sitting here today, it's sometimes hard to forget that a few years ago in 2019, there was a real concern about inflation being too low and the Fed not having the tools to fight low inflation. That gets to this low interest rate world where in the teens, the average short-term interest rates were a couple of percentage points lower than they had been, say, in the '80s and '90s. So that meant that monetary policy had less room to fight recessions because the Fed has decided it doesn't want interest rates to go below zero. And at some point, interest rates can't go infinitely negative. So at some point, there's a limit to, in a lowest interest rate world, how much a central bank can fight recessions. So if you conducted policy ignoring that, in other words, say, using a Taylor rule that didn't take account of the zero lower bound, it's easy to show in model simulations that that leads to a downward bias to inflation. So my colleagues, Dave Reifschneider and John Williams, who's now the New York Fed President, showed this a really long time ago. And others had shown it as well. So that's a problem, right?

Roberts: The central bank needs to do something when average interest rates are low to try to better achieve its congressional mandate of maximum employment and price stability. So what Mike and I did in our Brookings paper was to explore a variety of these different approaches. They all have this flavor of... They're called sometimes lower-for-longer policies or makeup policies. And the basic idea is to have lower interest rates in the future than a conventional policy would suggest. So why is that a good thing? Well, this gets back to this idea that spending, like consumer spending and investment, react more to longer term interest rates than they do to short-term interest rates.

Roberts: And then, again, for the economists in the audience, if you are familiar with the expectations hypothesis of the yield curve, then you would know that expectations of short-term interest rates in the future being lower would make today's long-term interest rates lower, and that could stimulate the economy today. So that's the idea of these lower-for-longer or makeup-type policies that you're lowering the interest rates in the future to make long-term interest rates… You promise to lower interest rates, keep them lower than you otherwise would in the future to make long-term interest rates lower today.

Beckworth: And you look at a number of rules that implement this, or approaches, I guess, that implement this. You do it in both the papers, right?

Roberts: Yes.

Beckworth: You take a look at both. So let me just jump to the second one with with Bernanke as well as Michael Kiley. I'm going to read from the table here, all the different rules you guys use. So these are monetary policy rules. You start with a Taylor rule, then you do an inertial Taylor rule, which inertial means it slowly adjusts. It doesn't automatically do everything that the model parameters say you would do. So you lag with your changes. Then you do a flexible price level target. You do an inertial flexible price level target. And I believe a flexible price level target is a price level target with the output gap. You adjust for the output gap. Is that right? And then you do a flexible temporary price level target, a temporary price level target, several temporary price level targets with different memories. So the average, I guess, three-year memory, one-year memory, the Reifschneider-Williams, and then the Kiley-Roberts change rule. So you go through a lot of rules. And is there any takeaway from all these rules? Any big insights?

Roberts: Big picture is that a lot of these rules like the one that John Williams and Dave Reifschneider came up with, most of these rules that you were talking about that promise lower interest rates in the future like a price level target or a temporary price level target, they're all better than just ignoring the issue. So the big message is you can do better by using a lower-for-longer policy than you could by not using such a policy.

Beckworth: So this is why I've been a fan of level targeting for some time. And I know these questions really first got considered back with the Japan deflation. Ben Bernanke, the whole Princeton group, Michael Woodford and all of those… Paul Krugman, his famous 1998 Brookings paper, Lars Svensson, they all were talking about how you escape from a deflationary environment. And level targeting is always a key part of that or makeup policy. And at least the intuitive understanding I take away from it is, if people believe that you're going to return to a level path and they take it and it's credible, they themselves will do some of the heavy lifting. It makes it easier for the monetary authority. If you expect inflation to go up and return the level to its path, the market itself, people themselves, if they believe it, they'll do some of the heavy lifting for the central bank. Is that right? They're thinking intertemporally. They'll spend more than they otherwise would if they know it's lower for longer.

Roberts: That's right. There are various ways of looking at the issue. And so I was emphasizing one way that… all the policies that you were just describing, those level-type policies, are promising lower interest rates in the future. So you can think of them through that lens that, "Oh, I see lower interest rates in the future meaning lower long-term interest rates today." But it's exactly the same mechanism. It's just thinking about it different ways. And then you bring up the additional feature which is that, if inflation is expected to be higher in the future, either because those interest rates are going to be lower in the future than they would otherwise have been or explicitly because you've got a price level target or a nominal GDP target where the Fed is announcing that they're going to be trying to have higher inflation in the future… either way you look at it, that's going to raise expected inflation in the future, which could ripple back to the present and raise inflation expectations today, which raises inflation both today and looking into the future, which means that real long-term interest rates are going to be lower than they would otherwise be.

Roberts: So one of the issues that we explored in the paper with Ben Bernanke is the importance of these two different aspects where one would be the lower nominal interest rates in the future lowering nominal long-term interest rates today versus this inflation expectations dimension. And so in the simplest models, they both matter. But there's some skepticism. I think a lot of economists will more readily accept that people in financial markets have this high degree of rationality when they're thinking about their expectations.

Roberts: If you think about all the Fed watchers who are trying to figure out exactly what the Fed is trying to do, it's pretty clear that they're paying attention to what the Fed's going to do in the future. It's less clear that all the businesses and households who are involved in setting wages and prices are thinking that carefully about what the Fed is going to do in the future. So it's an important and interesting question, which of those two mechanisms is the more important? And what we find in the paper with Ben is that if you shut down the inflation part of it, the policies still work pretty well. So just relying on the rationality of the financial markets gets you a lot of what you're trying to accomplish with the lower-for-longer policies.

Beckworth: I've often been asked, "David, why would a household, why would the public care or even know about a level target, price level target, nominal GDP level target?" And my answer has been they don't have to know explicitly about it. All they need to know is the symptoms of it that they see around them. So you're right. If businesses, if Wall Street, if they're thinking carefully and they see it and they make decisions, they hire more people. Incomes are flowing more readily. Households, they just need to see that part of it. They don't need to be fully thinking about the 30-year projection of the Treasury yield. I think you can tell a story where even households, if they don't have that long view, can still benefit and be a part of that story.

Roberts: I think that there can be an effect of speeches by Federal Reserve officials on the public, not all the time and not for fine-tuning stuff. But if you look at the Volcker disinflation, that probably happened more rapidly than could be explained.

Beckworth: That's a good example.

Roberts: So there is some effect. The problem is it's probably not as crisp and clean as in some of those models that rely on a high-degree of rationality.

Beckworth: Well, I'll give an example where a policymaker's words were incorporated by the public and that would be during the Great Depression, FDR’s fireside chats, Gauti Eggertsson’s well-known paper. He shows people were listening. Of course, it's a very dire strait. It's a very different world. But that's an extreme example that illustrates the point. Maybe if you go to the other extreme, a case where things didn't work out, and Michael Woodford talks about this and some of his work is Japan. They did QE in 2001 and 2006, and it really didn't do a whole lot. He looked at it in many different dimensions. One way he looked at it was through the monetary base. He showed there was basically a temporary increase in monetary base, and it's completely unwound. And he goes, it's as if the public understood this was not going to be lower for longer or a permanent increase in government liabilities, and so it wasn't very credible. I think that's a key part of your work. It's got to be a credible policy for this to work.

Roberts: Absolutely. And this also gets to a little bit of behavioral economics-type issues. Economists like to talk about a concept of rational inattention. So this is the notion that for every press conference, people aren't going to be paying attention to every press conference that Chair Powell gives. It's just not worth their while. But for really big things like the Volcker disinflation, that will be on people's radar screen and maybe more likely to pay attention. I think your story about Japan could well be right that people weren't fully convinced that the Bank of Japan was going to follow through.

Beckworth: Yeah. Well, John, let's transition out of this discussion about the zero lower bound, really fascinating. I could talk to you all day about this. And let's go to an application of it, and that is the Fed's new framework, FAIT. So in your paper, you talk about temporary price level targets, and Rich Clarida is, as you mentioned, vice-chair. He had several speeches where he said FAIT is a form of a temporary price level target. So what is your take on FAIT? Is it a temporary price level target? You've expressed to me previously you're not excited about the name FAIT too. So maybe walk us through these issues.

Evaluating the Nature of FAIT

Roberts: Let me back up a couple of steps and say for the first time in 2019 and 2020, the Fed had a formal policy review of monetary policy. The intention is now to do it every five years. So they would start another one in 2024 which, I guess, now, is next year. For that 2019 process, a big concern was the low interest rate environment and how should policy be adjusted to take account of that. So what they did, they made changes in the summer of 2020 in a couple of different ways. So one is, there's a constitutional document that the Fed has, their statement of longer run goals, I believe, they call it. And they made a number of changes to that document… one aspect of it did sound quite a lot like temporary average inflation targeting. So they were very concerned about hitting the effective lower bound. And they said that what they would do is to make up any shortfall of inflation while interest rates were at the lower bound in the future, so definitely a kind of short run average inflation targeting. So they definitely introduced that.

Roberts: They also, in that document, introduced the concept that they would be making up shortfalls in employment for maximum employment. But they wouldn't push back against employment that was higher than their estimate of maximum employment, so putting in those two asymmetries on both legs of the dual mandate. So that's their constitutional document and definitely has a flavor of average inflation targeting. Then in September 2020, they implemented some of these things for the first time. So remember in the summer of 2020, monetary policy was very loose. The federal funds rate was at zero. And remember in July 2020, the unemployment rate was 10%. Inflation was 1%. So it was very sensible at that point to be at the lower bound. So at that point, they implemented their new overall framework by introducing some thresholds to the federal funds rate that were effectively lower-for-longer policies.

Roberts: So they said that they would keep the funds rate at zero until the unemployment rate reached their estimate of maximum employment. They didn't spell that out, but that could well have been, say, 3.5% unemployment because that's where they were pre-COVID and that was not inflationary. And they said that the other condition for raising interest rates above zero would be that it would look like inflation was eventually going to overshoot enough to make up the shortfall that was occurring because of the COVID recession. So that was not saying… the threshold wasn't that the average inflation had already reached 2%, but that they were expecting it to. That's very much an element of average inflation targeting.

Roberts: So at the time, based on their experience in the late teens, they had thought that the hard one was going to be the inflation part of it and that they would be hitting the unemployment trigger first. And then sometime down the road, the inflation trigger would be hit. In the event, it turns out that by the middle of 2021, the shortfall in inflation had already been made up, not just its forecast to be made up, but it had already been made up. But the unemployment rate was still, I think, around 5%. So they had had this dual trigger, but it didn't quite work out the way they had been expecting. So they did this framework review, and they first implemented it right away because of the COVID economy.

Beckworth: John, let me go back to the detail about the maximum employment criteria, and it says shortfall from maximum employment. Prior, it said deviation. So that means it’s symmetric. If it's above maximum employment or below, they would respond. So what do you think of that? Do you think that's an appropriate way to do policy? And what comes to my mind is something like Milton Friedman's plucking model when I hear just about shortfalls from maximum. That plucking model, Milton Friedman said, "The economy will go below potential but rarely goes above potential." Were they thinking that? Or was this some other motivation for looking just at shortfalls?

Shortfalls and the Symmetry of FAIT

Roberts: Right. So first of all, looking at shortfalls is actually, in my view, more consistent with their mandate from Congress. Their mandate is to achieve maximum employment. And if you go above maximum employment, either your estimate of maximum employment is wrong or you'll have an inflation problem in which case the other leg of the dual mandate tells you to tighten monetary policy. So you don't need to have conditions that say that you need to tighten monetary policy.

Beckworth: It's implicit.

Roberts: Right. So that that's one aspect of it. The other is that, remember, an analysis of monetary policy in the teens, and I think you had Joe Gagnon on who was making this point, was that they kept thinking that inflation was just around the corner because unemployment was getting below estimates of where they thought the natural rate was. And so they would tighten monetary policy, but then a few months later or a little bit later, they would regret it realizing, "Oh no, we can actually go lower on the unemployment rate than we thought." So I think that's another reason why they went that way.

Beckworth: So if you're going to err, err on the side of being more forgiving than being too tight and too severe...

Roberts: Based on their experience. And remember that the price stability leg is a check on that. So as long as you've got the price stability leg, you don't have to worry about being too generous on maximum employment.

Beckworth: Again, it's implicit. If you're going to aim for price stability over the medium or long run, you're going to, in some fashion, address it.

Roberts: You can't, for a long time, exceed maximum employment.

Beckworth: Right. Well, the time we have left and in closing, let's talk about where FAIT should go. So as you mentioned, their framework review is coming up next year. It's hard to believe it's next year, 2024. And then I believe they make a decision in 2025, so they have a year long process of reviewing it. Is that right?

Roberts: Yeah, exactly. My guess is they'd probably aim for early 2025.

Beckworth: So if you could wave your magic wand, how would you tweak or adjust FAIT?

Adjusting the Structure of FAIT

Roberts: So I think in that statement of longer-run goals, that constitutional document, I would keep, for the reasons I was just saying, the shortfalls language on maximum employment. But I'm not sure I would keep the shortfalls language on the average inflation targeting. My reasoning is that, if you think about it, that puts an upward bias to inflation. When they adopted that, it seemed like it was a good idea to have an upward bias on inflation, because their concern was getting stuck in a low inflation world because of the low interest rates. Today, it doesn't look like we have a problem with inflation never being able to be high. It looks like that's perfectly possible. So I don't think, from today's perspective, they need that asymmetry in the constitutional document anymore.

Roberts: And then in terms of if they ever do hit the lower bound again, what statement language they should use… So I think in retrospect, it looks like having those two criteria for liftoff, both the maximum employment and the price level-related one, led to monetary policy being too loose for too long. So a really simple change in my view would be to change the "and" in that conditionality to an "or." So you would say then that, "We will keep the federal funds rate at the effective lower bound until either we've achieved maximum employment or some kind of average inflation targeting criterion has been reached." So that would've stood them in good stead this last time around, because by the middle of 2021, they had already made up the shortfall in prices. So that promise would not have gotten in the way this last time. And it would've stood them in good stead in the previous cycle where it took a long time to get to maximum employment and inflation kept undershooting. So it would work for each of the last two cycles. Who knows? Maybe some other thing will come up that would make it not work in the future, but at least, it would have worked, I think, in the last couple of times. So the tweak there would simply be to change "and" to "or."

Beckworth: And make it robust to different situations.

Roberts: That would make it more robust to different situations.

Beckworth: Okay, well, with that, our time is up. Our guest today has been John Roberts. He blogs at John Roberts Macro. We have a link at the bottom of the show notes. John, thank you for coming on the show.

Roberts: Thanks. It's been fun.