Tara Sinclair on Real-time Economic Analysis and the Fed’s Upcoming Framework Review

Tara Sinclair is a professor of economics and international affairs at George Washington University, where she also directs the George Washington Center for Economic Research. From 2022 to 2024, Tara also served as the Deputy Assistant Secretary for Macroeconomics in the Office of Economic Policy at the US Department of Treasury. Tara joins David on Macro Musings to talk about her time at Treasury, real-time economic analysis, the Fed framework review, and much more.

Check out our new AI chatbot: the Macro Musebot!

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: Tara, welcome to the program.

Tara Sinclair: It's so great to be here.

Beckworth: It's great to have you on. Now, we've met before a few times. In fact, I have had some visits to Treasury when you were there. You were a DAS, as they call you, at Treasury. I know some of the people that worked for you— Alex Schibuola, Andrew Martinez. I met some of your other colleagues as well, but that was a great office there that you had.

Sinclair: It's a fantastic team of people. I deeply miss working with them. I learned so much from that team.

Beckworth: Well, tell us about that experience. What is the role of a DAS covering macroeconomics in the US Treasury Department?

Working as a Deputy Assistant Secretary at the Treasury Department

Sinclair: Well, first of all, I will say that it was a particularly unique time being at Treasury having Janet Yellen at the helm. That was really special, but the team has been there for a long time doing really important work, and one of the key things that I learned from being a political appointee there is the importance of the longtime career staff and the amazing work that they're doing. And so, they're there making sure there's continuity of deep understanding of macroeconomic developments, and then they're passing that up to the politicals who are coming through, in and out, and were moving at a completely different pace than they're moving, because they're able to think deeply about things. And hopefully, they're thinking deeply about a broad range of things, so that when something comes up and I say, "I need an answer in an hour," I can get up to speed really quickly from the team. That's exactly what I got from the macroeconomic analysis team at Treasury. They're a fantastic group of people.

Beckworth: You raise a great point there. One of the big differences between being an academic and being in a policy shop, or even in the think tank world, but particularly in government, is that you’ve got to have analysis done ASAP. I want to know in an hour, 30 minutes, give me a memo. You’ve got to really be able to crank that out. Unlike an academic, [where] you can sit back and think for months, write a paper, go to a conference.

Sinclair: Right. Well, and oftentimes, the events that happen are exactly the things that, at least in my case, I'd never thought about before. For one example, I was there during the banking turmoil of March of 2023, and a lot of that happened over a couple of weekends. And so, I had to quickly learn everything about the financial system and the role of banks within the financial system, which was very far from my academic research. Therefore, I really leaned, both on the team, as well as all of the other research that people were putting out. There was some great stuff that academic researchers were putting out very quickly as well that helped me, but it was a crazy pace and a very different role where I wasn't choosing the topic, [and] the topic was just being dropped in my lap.

Beckworth: So, you had to respond to the events as they unfolded, but you also had some normal routines. So, walk us through the day in the life of a DAS for macro at Treasury.

Sinclair: So, I used to always say that there is no typical day for a DAS. However, mine did have some structure to it, which started at the beginning of the day, where every day we had what ended up being a 9:00 AM meeting that we typically call the markets meeting, and we would get updates from everything that happened on international markets overnight. But I also pigeoned my way into a speaking role to talk about the macroeconomic data that was typically released almost every morning at 8:30. Macro data comes out— It's relentless. It comes out almost every morning at 8:30, there's something new to talk about.

Sinclair: And so, we would have this meeting, and it was organized by the deputy secretary, but the secretary, Janet Yellen herself, often attended as well, and most of the major leadership of the building was there. It was a quick Zoom meeting. You might want to think about it as a standup meeting from tech firms. It felt very familiar to me from that space, and we would get a quick update on macro data and on the markets. Then, that really set the tone for the building that day.

Beckworth: I imagine that your briefings became increasingly important because, as we learned, suddenly the CPI reports became a whole lot more important than they were in the past, the labor reports. Those are always important, but even more so with the inflation surge.

Sinclair: Right. Well, in particular, we were really carefully watching, as the data evolved, to get a sense of where the Fed was going to go, because there isn't direct coordination between the Treasury and the Fed, and so we're watching what they're doing because we do want to have a reasonable macroeconomic policy, and we want to understand how the Fed is thinking about things so that we can incorporate that in the way that we're thinking about things as well.

Beckworth: So, would the Office of Debt Management be a part of that meeting as well? Did they say, "Hey, we need to reprice our debt when we issue new debt and we auction debt"? How did that work?

Sinclair: No. They didn't speak, typically, in that meeting. They had representatives at that meeting, but I will just say that my sense of what they were doing was much more focused on longer-run goals. One of their clear views on policy, from a debt management perspective, is that they want to be consistent and predictable, and so, they weren't making small response to data-type changes in any of their policies.

Beckworth: Alright, very nice. Sounds like a great experience there, and you had a lot of interactions with Secretary Yellen at a very important time. We'll talk about that in a minute, some of the data issues you had, in particular. But any lessons or takeaways from that, particularly with [respect] to data? You've been there, you see what the government needs. If there is a wishlist, if you could wave a magic wand and get the data that you wanted— more real-time data or maybe better data than we already have, better versions of it— what would you call for?

Building a Better Economic Database

Sinclair: Well, if I really had a wishlist, I'd like to be tracking every single action of every single economic actor and keeping that all in a nice clean database. We're so very, very far from that. We're much farther from that than I think a lot of people even think, given the amount of private sector data that is being collected, because it's such complicated administrative data, [so] to try and actually connect it together is really, really challenging. And so, I particularly admire the Bureau of Economic Analysis for the work that they do, [for] putting all of these complicated data sources together.

Sinclair: But I think, right now, we're so far from that wave-a-magic wand wishlist, because we're far from where we've even been in the past. We're looking at cases where our statistical agencies, actually, in real terms, have a smaller budget than they have historically, and they're facing greater and greater challenges for collecting data. We've got lower response rates, people aren't picking up the phone anymore, people don't want to respond to surveys, there's survey fatigue, and we've got a larger population, we've got more people to track.

Beckworth: That's interesting. So, it's not just that Congress isn't funding the agencies, it's that people themselves aren't responding to the surveys.

Sinclair: Exactly.

Beckworth: So, on both sides, you're having challenges. I want to bring up someone who used to be my boss, Bill Beach. He was actually the head of the research program here at the Mercatus Center. He became the BLS commissioner, and he went on Twitter and really made the case for more funding for the CPS sample size. So, maybe speak to that issue.

Sinclair: Yes. This is a really dire situation, and I really appreciate that many past commissioners are stepping up and making this sort of call for re-funding of the CPS, in particular, because it is already at the point now where they've decided to cut back on the sample size because they just don't have the staffing or the funding. And the way that the funding works, it's complicated, because— you might wonder if they can pull money from other surveys or something like that, but there's specific allocations for different types of data releases.

Sinclair: So, for the CPS in particular, they're constrained as to what they can do, and at this point, it seems like their only option is to reduce the sample size. And this is precisely when we really want to know what's going on with the labor market, [when] we particularly want to track what's going on with the unemployment rate, which comes from that survey, and yet they just don't have the funding to do the amount of surveys that they have to do now in order to get up to the sample size that they've historically had.

Beckworth: So, the CPS— that has the unemployment rate. Does it have any other important data measures in it that we would care about?

Sinclair: Oh, it has the employment-to-population ratio, the prime age employment-to-population ratio, which a lot of people really care about, and then all of the demographic breakdowns that we want to track and understand. When the Fed talks, for example, about thinking about a broad-based and inclusive employment target, how are they going to do that if they don't know what's going on with the underlying demographics?

Beckworth: It's getting harder and harder to even think about that. So, one of the issues that comes up in this discussion is real-time data, and I have this vision of real-time data that— man, you guys have access to satellite images of ports, of factories, of lights going on, or there's credit card transactions, or some measure of traffic. It's a whole new world, but it may not be that ideal, as I make it out to be. And you have a speech on this very topic that's coming up at the Bank of Canada, and it's on this issue of real-time data analysis. Before we get into your speech though, maybe define for us, what is real-time economics, real-time data?

Breaking Down Real-time Economics and Real-time Data

Sinclair: The way I think of real-time economics— It's very old school. We can get into the private sector data that you mentioned in a minute, because I think that's an important future aspect. But one of the things that we really face when we think about real-time data is that the statistical agencies that are putting out the official government data— they face a trade-off between timeliness and accuracy. And so, in order to get us timely data, they often, for many data releases, have an initial advance or preliminary release of the data. Then, as additional surveys come in or more information from other data sources come in, they revise the data and provide us new updates. We see this with GDP, we see this with the employment numbers, and this is a normal, globally accepted policy of how we do statistical data releases.

Sinclair: And so, the reason that they're releasing that timely data is for decision makers that need to make a decision today— they can at least have some insight as to where the economy is going, and they can make their decisions today. Yet, a lot of times, when we do economic models, we look at the data as it's been revised. And so, [with] real-time economics, we take seriously that the data often faces revision processes, and so we want to think about decisions and how they're being made in real-time with the data that's available from those initial preliminary releases.

Beckworth: That's a challenge for policymakers, right, if you're at the Fed, if you're looking at GDP? I think, if I remember correctly, [in] 2008, it wasn't looking too bad at first, and then [after] revisions, it was actually pretty bad.

Sinclair: Exactly.

Beckworth: But is that the reason that the Fed, for example, often looks at labor markets? I know that it's part of their mandate, too— they have maximum employment— but just practically, they can look at labor markets more monthly. It's accessible.

Sinclair: Yes, so, the CPS data that we were talking about earlier— for the unemployment rate— that is only revised for seasonal factors, and historically, even real-time data economists have been like, "Oh, it's only seasonal factors." Although, it ends up that, for CPI, another survey that is only revised for seasonal factors— that one did see some pretty big changes in 2022. So, watching some data sources that are not revised might be one way to prevent some of these real-time data issues. But I think that modeling the revisions themselves and keeping that in mind and maybe thinking more broadly about decisions in this framework of broader uncertainty is important for policymakers as well.

Beckworth: Let's talk briefly about the private sector data that I referenced, and also, maybe nowcasting, because that's a cool, sexy thing now. The Atlanta Fed has a GDPNow measure. Everyone seems to point to that. Would you say that that one has won the competition of who's got the best nowcast? Is anybody else competing with them?

Sinclair: Well, it's interesting, because they've definitely gotten a lot of attention, but oftentimes they're overly optimistic. And so, there are a range of other nowcasts that are out there. Most of the regional Feds put out one or another, and it's fun. So, I tend to like to watch both the Atlanta Fed’s GDPNow cast, and then the St. Louis Fed’s, because that tends to be the whole range. The St. Louis Fed tends to be negative and the Atlanta--

Beckworth: Upper bound, lower bound.

Sinclair: Exactly. Exactly. That's an approximation, but it often works pretty well, just in a pinch. But one thing in past research that I did, well before I went to Treasury, [is that] I actually looked at the different performances between forecasts— so, thinking literally about a quarter ahead or more, versus being able to predict the current quarter, either doing a current quarter forecast or using one of these complicated nowcast models that's bringing in data as it arrives. And it really seems like there's just a huge difference in our ability to understand where we are versus predict where we're going. And so, in general, I think that it makes a lot of sense to be focused on nowcast models, both because they're much more informative than forecasts, but also, we have to get those right as a jumping-off point to have any possibility of getting our forecasts right.

Beckworth: That's very interesting. So, one other measure that I've been watching lately, and I'd love to hear your thoughts on it— the New York Fed started this measure, and I believe that it's now carried by the Dallas Fed, but it's the weekly economic indicator, which is pretty wild if you think about it. [It’s] a weekly update to a measure that's supposed to approximate real GDP, and it's fun to watch it. If you go back and look— and I know there's revisions to that as well, because inputs that are fed in have to be revised— but the one big drawback to it is that it's on a year-over-year basis. It's literally like a percent- change from a year ago. So, you can't really see quarter-over-quarter or maybe even week-over-week in the same sense that you would a real GDP measure.

Sinclair: Yes. Well, it's interesting. I actually talked with Jim Stock about that very issue, because he was one of the developers of it, initially. And he even pointed out that what we really need is an effective seasonal adjustment for weekly data. That's why they're doing the year-over-year. It's for seasonal fluctuations. And there's just so many things that happen at the weekly pattern that change but that have that seasonality to them, that we really just don't have a way of seasonally adjusting those in any way that they found that worked better.

Beckworth: That's interesting. So, in principle, they could turn the weekly economic indicator into a truly comparable measure to real GDP, but they'd have to first deal with the seasonality issues. So, maybe, maybe, AI is a solution that's going to solve this at some point in the future. So, I'll hang my hope on that.

Sinclair: Maybe.

Beckworth: Let's go to your speech, though. Enough of this discussion about nowcasting and other things. You have a speech coming up at the Bank of Canada. Tell us about it.

*Real Time Economics: Tales from the Trenches*

Sinclair: I'll be speaking to the Real-time Economics Conference. So, this is the RTD or real-time data conference. It's been going on since about 2001. And so, it's a group that I've been involved with since near their beginning. It's such an inspirational group of people that thinks very deeply about these data issues and how they enter macro models, and I was, admittedly, so inspired by them early on in my career, and it's such an honor to be able to go back and be their keynote this year. So, really, what I wanted to share with them was all of the ways that things that I had learned from my interactions with that organization over the years played into my work at Treasury, and there were a lot of stories to be told.

Sinclair: One of the key ones that I highlight is what happened with these seasonal factor revisions to CPI at the beginning of 2023, so looking back at the 2022 pattern. You might remember that CPI inflation peaked in mid-2022, and so then we were all carefully watching to see exactly what the pattern was going to be coming down, and a lot of the year-over-year measures were hard to track because they were reflecting things that happened a year ago. And so, everybody just started really watching the monthly data. And so, for the monthly data, we have to use the seasonally adjusted data. There, we're watching those month-to-month changes, and then sometimes people would [say], "Okay, let's look at three-month averages to at least smooth out a little bit of that," but what we initially saw was a pretty clear trend that the end of 2022 is really where we saw a pretty strong disinflation.

Sinclair: Then, in February, the Bureau of Labor Statistics released the seasonal adjustments, and it changed the whole pattern to where, then, it looked like there was less disinflation at the end of the year and they just reallocated, because it's just what happened within the year. The year-over-year seasonally unadjusted data stayed the same, but we were all watching so carefully, that seasonally adjusted data, that it really changed people's expectations. And this was right in there. So, Chris Waller made a speech highlighting the last three months of the year and the improvements that he was seeing, and the Fed made a change and reduced— they went [with] a 25 basis point cut after they had done 50 basis points before. So, this was all really playing into the policy discussion at that time.

Beckworth: That's amazing. So, these seasonal factors really were the statistical agencies pulling the rug out from under the Fed, "Sorry."

Sinclair: Right.

Beckworth: Now, is it the case that seasonal factors are harder to estimate in the midst of a big crisis?

Sinclair: Exactly.

Sinclair: I definitely do not envy the statistical agencies in any of this time period for all sorts of reasons, but particularly for being able to update these seasonal factors. Because they do evolve over time, so they want to go back and use recent data to update those seasonal patterns. But then, what about the patterns in COVID, and even in the years after COVID, were seasonal patterns that we should be using going forward, versus which is about the waves of the virus? There were all sorts of challenges there for them to try and come up with appropriate adjustments.

Beckworth: Yes, so, big changes during the pandemic. One that comes to mind is the shift between services and durable goods, for example. I'm sure that that affected seasonal measures and stuff, so, lots of noise, lots of having to go back and re-estimate. Now, you also highlighted, in that speech, that there were other measures that also were revised due to seasonal factors. Maybe speak to some of those as well.

Sinclair: Yes, and revised due to all sorts of other data revisions. So, one of the particular challenges that we faced in September of '23 was that we got the five yearly comprehensive updates for GDP and other NIPA account measures from the Bureau of Economic Analysis, and there were at least three major stories that we had been tracking in terms of macroeconomic events. We were tracking excess savings, we were tracking the relationship between GDP and GDI, as well as tracking the role of the labor share of income.

Sinclair: And these all matter for thinking about things like forecasting where the federal budget is going to go, and for understanding the economic experiences of different people. And also, with excess savings, just forecasting where the economy is going to be going in terms of the impetus of further consumer demand. And all of those shifted dramatically when we got the new updated data from the Bureau of Economic Analysis. And so, [with] every chart, every explanation, we had to make sure that all of the leadership in the building understood that any of their old talking points needed to be updated because of this new data.

Beckworth: So, just stepping back, we have all of these challenges with this data. It gets updated. As you mentioned, there's an initial release and we get more information. Throw, on top of that, seasonal factors, [and] throw, on top of that, that people don't respond to surveys. There's just a lot of struggles, it sounds like. Do you see the advent of big data, of maybe AI, offsetting any of that going forward?

Solving Our Current Data Challenges

Sinclair: I'm so glad that you came back to this, because I spent 10 years working with Indeed, the job search site, and my initial goal when I went in there— I guess it was about 12 years ago now— I thought we were going to save the world with all of this new private sector data, and I think that it's really important and really useful, but it's definitely complimentary to these long time series of data that we're getting from the government, and we do need to be able to ask these targeted survey questions and get specific answers that we're not going to get from just tracking all of this different private sector data.

Sinclair: And so, of course, I think that there's value in AI and in these different sources of data, but the complexities of what private sector companies are trying to do, and trying to align that with the information that the government wants— It's way harder than I thought when I first got into it, than a lot of people thought. I was regularly talking with policymakers during my time at Indeed about, "Oh, just give us the data," and I'm like, "There isn't ‘the data.’" It's a giant mess of various pieces of information that may be relevant in certain cases, but it takes a lot of work and then businesses change. They change what they collect.

Sinclair: And so, counting on that— this is something that statistical agencies have found, that even when they start bringing in some of this private sector data to enhance the data products that they're building, then what happens if they can't renew the contract? What if Congress doesn't give them the money to pay for the new contract? What if the company goes out of business or changes their business model and doesn't provide that data series anymore? We have to have the central statistical agencies.

Beckworth: So, even to the extent that they do bring new added value to the data process, they themselves are very uncertain, in the future, that they'll provide data that they can change. So, there's a whole set of new problems that come with this private sector data. So, it's not as simple as in my little dream world of satellites looking down on planet Earth and train traffic and--

Sinclair: Even with the satellites, patterns of light use have changed, and so, being able to extract information by the time we've modeled it and said, "Okay, this is going to tell us about GDP—" Patterns change, and it's not as informative anymore. That's something that we've seen again and again.

Beckworth: So, big takeaway, have some humility about what we can accomplish with our data, but still pursue it and push Congress to fund it. That's probably the biggest area for improvement right there. Let's go from data and real-time analysis and focus in on the labor market a little bit more. You recently gave remarks as a former Treasury official. You're getting lots of invitations— Bank of Canada, the National Association of Business Economists. You gave a talk on labor shortages and job mismatch. So, walk us through that presentation.

*Labor Shortages and Job Mismatch*

Sinclair: Sure. That's actually related to a paper that I have that's joint with some co-authors at Indeed and at the OECD. What we're looking at is a set of OECD countries, and we've got Indeed data that is pretty comparable across the different economies, which is one of the exciting things about this particular private sector dataset, is that we can do a comparison across countries that's all collected by the same institution. What we're wanting to look at— and this is something that I've been thinking about for a long time when looking at the Indeed data— is that, in aggregate, we always think about the V-to-U ratios. We think about the vacancies and the unemployed as if we're just going to take any unemployed person and drop it into any vacancy, and that if it's a one-for-one, then we're in some kind of stable situation, but in reality, people are very different. Jobs are very different.

Sinclair: And so, I think another structural issue that we need to think about is what researchers have called labor market mismatch, which is where the— and you can do it by geography or by industry or by occupation. In our case, we're looking at an occupation-type mismatch. So, we're thinking about the kinds of jobs that people are looking for, which, I think, is a combination of what jobs they think they can get and what jobs they're interested in. I tend to emphasize both their skills and their interests on the job seeker side. Then, on the firm side, obviously, they have particular roles that they need to fill, and so they're looking for certain skills and interests. And in many cases, there are jobs out there where there's 100 applicants for one job. Then, there are other jobs where they struggle to get any applicants at all, which suggests that there's a lot of mismatch in the labor market and that we could get some improvements in the economy if we just brought those two groups together more.

Sinclair: Primarily, we're not going to get employers to change the kinds of jobs that they're offering, but we can perhaps make them more appealing. So, traditionally, economists think through wages, but also through features of the job that aren't necessarily the day-to-day tasks that they need done, but could they be done remotely? Could they be done in a more comfortable environment with snacks? That sort of thing. Then, of course, on the job seeker side, people often talk about training and development and, I think, just information sharing, so that people know what kinds of job opportunities are out there. And so, my talk at NABE was really focused on— we were talking about different aspects of stimulating labor demand and labor supply and bringing them together, and I wanted to highlight this other feature of the labor market that we need to think about as well. That our labor supply is not a monolith of same-agent type people. It's very different people, and we need to think about how to match them to the jobs that are out there.

Beckworth: Again, going back to the pandemic, did that disrupt or throw a new kink in the production process of bringing labor demand and supply together?

Sinclair: It did. And so, what's interesting is that we initially saw— across the OECD countries, we saw a big spike in mismatch. Then, we saw a pattern where the countries that had some kind of job retention scheme policy, where they were really trying to keep their workers attached to their employers— those countries actually saw a slower improvement on their mismatch metrics than countries, like the US and Canada, where people went onto the unemployment rolls, got unemployment insurance benefits, and then went back and re-matched with jobs after the pandemic.

Sinclair: And, of course, lots of them went back to their same past job. There were tons of temporary layoffs, but there were also a lot of people that went and found new, better jobs and actually took advantage of some of the long-term structural reallocation that was happening in the economy. And so, my takeaway from that paper is that even though we think, traditionally, that the macro policy had been to preserve employer-employee matches as much as possible in transitory shocks— because that should be the fastest way to economic recovery. If there are these underlying structural shocks happening, it might be better to let them separate and find new, better matches, and that's what seems to have happened out of the pandemic.

Beckworth: Now, one other thing that you bring up in this talk, if I remember correctly, is that there's also some concern about the future of the labor force and the aging of it and the lack of growth in the supply of labor. Is that fair?

Sinclair: Yes. And so, there's two sides of the debate. There's the AI view that AI is going to take all of our jobs and so there won't be any jobs in the future. I am bullish on AI but not that way. I think that our jobs are possibly going to be quite different in the future, but I think that there are going to still be jobs. And so, what we want to think about is, will there be workers for those jobs? Both from an aggregate sense, which we have to be worried about because of the aging labor force, but also, will workers be interested in those types of jobs? Will they be well-trained for those types of jobs? And I think that's something that is an additional policy challenge that we need to keep an eye on.

Beckworth: Well, let's switch gears here. You are a macroeconomist at heart, even though you really love all of these other issues. Actually, these are the important issues to doing good macroeconomics. I come, I guess, from a background where we just played with the aggregate data, but you're more careful, more precise than some of us macroeconomists. But with that said, we both care deeply about the Federal Reserve and how it does its business.

Beckworth: And this year, here soon, the Fed is going to start its next framework review. Let me read from its Statement on Longer-Run Goals and Monetary Policy Strategy, also called the consensus statement— the very last paragraph, which I believe was put in in 2020 when they did the last framework review. It says, "The committee intends to review these principles and to make adjustments, as appropriate, at its annual organization meeting each January," so there, they're just going to renew and approve, or maybe tweak, "and to undertake, every five years, a thorough public review of its monetary policy strategy, tools, and communication practices.”

Beckworth: That last part, their review of the strategy, tools, and communication practices— that's the Fed framework review, and we've talked a lot about it on the show. I think that people know where I stand on this. I would love, one day— I know it's not happening this time around— if they ended up at a nominal GDP level target, but I would say that, maybe, at a minimum, let's move in that direction, even if we're not going to get there. But one thing that I've observed, I guess, in looking at all of the conferences and discussions that have been going on— I want to get your sense on this— is that, like last time, the review is probably going to focus mostly on the strategy part, not the tools and communications. Is that your sense, too?

Breaking Down the Upcoming Fed Framework Review

Sinclair: That is my sense. I think that there will be some discussion of communication. There have been some recent issues with communication. So, I think that that's probably going to be coming more to the forefront than maybe they initially thought even a few months ago as they started planning out the review discussion. But I do think that it seems like they're very focused on the strategy side, and it seems that most of the tools are off the table, which I think is really unfortunate. I would like to have a more robust discussion of the tools.

Sinclair: I think one of the reasons why tools are off the table is because this is an internal review, and I'm actually working on a paper with some co-authors comparing reviews of central banks around the world to understand how these different framework reviews have been done, and they've been done very differently. In some countries, they've had external reviewers with a broader range of expertise come in and provide recommendations. I wish that the Fed were doing that here. I know that there's going to be 50 shadow reviews, at least. So, there will be some of that. But to have the Fed have to face those recommendations and comment on them, I think, would be incredibly valuable.

Beckworth: That's a great idea, yes.

Sinclair: But instead, I think that they've said, “These are the tools that we know at this point.” And so, perhaps getting a bit more of a precise estimate about how QE works— some of that stuff will probably be discussed internally. But in terms of a big shift, I would like them to even go back to some of the operational--

Beckworth: Oh, hear, hear. You're preaching to me. I love it. No, I'm with you on that. That would be great to have, an outside evaluation, and just to suggest things that are very constructive. My colleague here, he's affiliated with this, Eric Leeper— He helped, I believe, on the Reserve Bank of Australia's review, and I think that Andy Levin also added to that. I recently had on Isabel Schnabel from the ECB, and they also did a review, but they did it of the operational system, which I think is great. I do think that those are big enough issues, [so] you probably want to do them separately. So, the ECB did adjust its floor versus corridor system, and I cared passionately about that, as listeners will know, but I wish that they would undertake that as well as the strategy review. But yes, those are great points. I wish that they would address tools, communication, and maybe they will. Maybe they will have some more comments on communication.

Sinclair: I think that they're going to have to have some comments on communication. Even just the most recent FOMC decision, when it seems like a lot of the action happened during the blackout period— It seems like there should be some further discussions about communication.

Beckworth: So, there's always room for improvement. All of us know this in our own life and the Fed [knows this] as well. So, let's talk about that strategy part, because that's where most of the action's going to be. And again, last time— that's where almost all of the action was. So, just to go back and maybe recap what we saw in 2020— We went from something called FIT to FAIT, a flexible inflation target to a flexible average inflation target, and let me just highlight the changes that were made. Two asymmetries were introduced. One would be the make-up inflation. So, when the Fed was below 2%, they'd make-up from below, but above, effectively, they went back to a FIT, or no make-up from above, so, make-up from below. 

Beckworth: Then, the other asymmetry was shortfalls from maximum employment, where before it was deviation, so above or below. So, it was symmetric before, now it's just shortfalls, it's make-up from below. Then, the third change, which I really hadn't paid much attention to until recently— and I'll tell you why. I think you know why. That is the definition of maximum employment being a broad and inclusive measure. So, Christina and David Romer had a speech or paper recently— I guess you attended that conference.

Sinclair: I did.

Beckworth: Where they said that this actually was the biggest, most important change, which surprised me and, I think, probably surprised a lot of people at the conference. But let's go back to the first two, because I think that the first two have received most of the attention and most of the suggested changes, if there are going to be some. Gauti Eggertsson and Don Kohn had a paper, I believe, in '23, a Brookings paper. They noted that this framework added, "An inflationary bias,” which, it was intentionally doing that because they were thinking of a zero lower bound environment. You want to get that— literally getting the average up to 2 from, say, 1.5, 1.6. But they also mentioned that that framework was not robust to all environments. In fact, I think they used the term, "It failed the stress test of the inflation surge." The Fed didn't do FAIT for very long. They ended up effectively abandoning it and going to worrying about inflation. So, you want something that can do both ways, right? High inflation, low inflation.

Sinclair: Right.

Beckworth: Mickey Levy [and] Charlie Plosser had a paper at the Hoover conference, and they made a point that's been stressed by many before, as well as at the Brookings conference, that I've seen. They highlight, particularly, the shortfalls from maximum employment as tying one hand behind the back of the Fed, and the argument is that you can't look at labor market indicators for the future of inflation because— only shortfalls. You can't look at, "Oh my goodness, it's overheating and there's going to be inflation in six months." So, that was their critique. Effectively, I guess, you could summarize it by saying that they were worried about the preemptive tightening not being on the table.

Beckworth: So, those are the two big criticisms. I'll mention one more actually. Michael Kiley had several papers he presented, I believe, at Brookings as well. He also emphasized concerns about the shortfalls approach. I believe that his argument was that if you focus too much on that, you can actually make the business cycle worse. You can cause bigger swings and end up with, actually, ironically, more shortfalls. So, I'm curious, where do you land among all of these concerns and issues about the Fed's framework?

Addressing the Concerns and Issues Surrounding the Fed’s Framework

Sinclair: Yes. Well, there's a lot there. So, I'll actually start out by talking a little bit about how the Fed does seem to traditionally fight the last war. And so, when they first announced, in 2019, that they were going to do the 2020 review, they highlighted that it was a quiescent time. “This seems like a good time to do a review of our policy, and we've had the same patterns for some time.” And so, that was really the framework that they had in mind when thinking about the framework was that, well, it seems like inflation is stable but too low, and these are the sorts of things that we need to be thinking about. And so, then, we got the framework review that we got.

Sinclair: Now, this time, they're definitely going to have the recent inflation period in their minds. So, in some sense, I think that this is going to be an opportunity for a dramatic course-correct. Although I do worry that, again, because they're only doing it internally, there might be some defensiveness there that, I'm sure, again, with the 50 shadow framework reviews— there will be criticisms, and then 50 more for years to come. But I do think that that's interesting, and I'm glad that they are doing another framework review with this new inflation context in mind, because that is half of their dual mandate, and, I think, really, their core responsibility in macro policy is the inflation side, and so to have a recent inflation episode in mind, in thinking about the framework, should be helpful. In terms of the shortfalls, I want to highlight this one, because, to me, I really think that this makes a lot of sense, because if you think about the dual mandate, it was maximum employment and stable prices.

Sinclair: So, from that perspective, maximum employment— it does sound like that one has an asymmetry there, where the prices has less of an asymmetry there. So, I actually think that those two pieces, both the flexible average inflation targeting and the shortfalls, aligns better with the dual mandate. I'm not sure that they're operationalizing that quite correctly, but I do see that it aligns well, and so, I think, in terms of communication, and, in particular, as policymakers, [they] do need to convey to the public the value that they're bringing. I do think that any suggestion by economists that too many jobs is a problem or too low of an unemployment rate is a problem— that's really just not great communication with the public.

Sinclair: But when we think about the role of tight labor markets today in predicting inflation in the future, it doesn't have a great track record. And so, I'm also not concerned about the shortfalls focus, from the perspective of that tying a hand behind their back or anything like that, because they're not going to do a particularly good job with it anyway. And I think that they could rely, and I think they do rely more and more on telescoping their inflation forecasts for taking policy decisions, and they can go straight to that argument rather than using the labor market in between.

Beckworth: So, in other words, don't rely too much on Phillips curves, which is fair. I have a lot of listeners who are like, “Phillips curves? Come on.” Of course, I just had some guests on who talked about, which version of the Philips curve— linear, non-linear? Is it the inflation expectations? Is it the output gap? Is it a supply shock term? Then, I have friends and listeners who are also big Monetarists [who say], "It's the money." I have my colleague, Eric Leeper [who says], "It's the future path of primary deficits." So, your point is that Phillips curve thinking— which is a critique, I guess— is it hasn't been the best tool. In theory, yes. Maybe in practice, not so much. It's very useful, maybe, retrospectively, to look back and try to maybe break down on things. Okay, fair enough.

Beckworth: Let's move, then, to this very, for me, surprising talk by the Romers. They argued broad, inclusive measure. Now, I thought it was actually interesting when it was introduced. It was a very politicized time. Maybe they were throwing a bone out for the FOMC. "Look, we do care," even if they weren't going to be that intentional in practice. But they said, "Hey, it actually did make a difference." They argued, to be precise, that it made a difference in the Fed falling behind the curve. You were there. Maybe you can give a better account. I'm not sure that they meant that it's necessarily a bad thing, but may have been the reason, in this particular case, that they were slow to respond.

Unpacking the Dual Mandate and the Fed’s Broad and Inclusive Goal

Sinclair: Yes. I think that they were really trying to find one piece of the revisions to the framework that could be connected in some way to falling behind the curve, and there was this sense, and I remember being part of this sense in real time, that we could possibly see a hotter labor market and not necessarily see inflation rising. And so, waiting a little bit longer to see that inflation rising— I did think that that made sense at the time. I don't know why the Fed waited as long as they did after inflation started rising. But I'm not sure— and I think that there's been other research papers recently that have said that if the Fed moved a few months sooner on that information, it would not necessarily have made a big difference in inflation outcomes. So, I didn't completely agree with their connection between those two things.

Sinclair: But what I thought was actually the hotter take on that was Caroline Hoxby, who came in and really spoke passionately about, if the Fed is going to be serious about this broad and inclusive measure, and that they're going to talk about, particularly, focusing on low-income, low-wage individuals and households, that they really need to expand their studies around these labor markets to understand how these people are actually experiencing work and financial life, because it can be very, very different than just saying, "Oh, well, they make lower wages." They have a much broader set of differences that need to be considered, and she really passionately highlighted that it doesn't look like the Fed is being very serious about that because of the lack of Fed working papers coming out on those sorts of topics.

Beckworth: Wow. So, she took the opposite view in a very strong direction. So, let's talk about the dual mandate, then, because, in my view, the framework review is part of a longer history. Yes, it started in 2020. Yes, the first consensus statement was 2012. But for me— I actually have a policy brief coming out on this— the framework review is just the Fed continuing to wrestle with how best to implement the dual mandate. It's remarkable that Congress did not define price stability or maximum employment. It also did not say how much weight to put on each one. It delegated a lot of power to the Fed to figure this out. So, to me, it's just this ongoing work— what is the best way to do it? And maybe we're learning as we go, and we're incorporating new insights from the profession, the academy, as we do it.

Beckworth: Price stability, we've now defined it at 2%. Jeff Lacker has an interesting paper that he wrote a few years ago. He goes through the history, because he was a part of the conversations in 2012, and you actually see conversations in the 1990s, within the Fed, [and] debates. He argues— and I've seen some empirical work that says that the Fed was effectively, or at least implicitly, targeting something closer to two by that point. But they were actually debating it, because it's not until Bernanke comes and really says, "We’ve got to go all the way. We’ve got to get past implicit and make it explicit." Jeff interestingly brings up the Summary of Economic Projections, which, I believe, started in 2007, but he said that in 2009, they started including the longer-run inflation projection which effectively was a target. So, Bernanke was very smart and careful, and Bernanke got a lot of support from Congress before he did this. So, kudos for him for getting the inflation target. So, we've got the 2% [target] and maybe people want to debate that, fine, but we've never defined maximum employment.

Beckworth: Let me go back to the consensus statement and just read what they say about it. It's a paragraph here that says, "The maximum level of employment is a broad-based and inclusive goal that is not directly measurable and changes over time owing largely to non-monetary factors that affect the structure and dynamics of the labor market. Consequently, it would not be appropriate to specify a fixed goal for employment; rather, the committee's policy decisions must be informed by assessments of the shortfalls of employment from its maximum level, recognizing that such assessments are necessarily uncertain and subject to revision." So, it strikes me that we'll never know what the maximum employment level is, and it's always going to be some kind of art, a little bit of “we'll see it when we see it.” How do you think about this?

Sinclair: So, again, when I think about the dual mandate, I think about it being, basically, can we run the labor market as hot as possible without going over? I do understand their 2% inflation target. I agree with you that maybe considering some other types of targets might be good, but they've said that inflation--

Beckworth: It’s not changing.

Sinclair: That's not going to change. Probably not even the 2% is going to change. There might be some discussion about it this time, but the 2% is not going to change. The target of inflation is not going to change. But if we think about inflation being where we want, get really close to that, but bring the labor market as hot as possible without causing inflation. I think that anything that the Fed can do that creates more jobs, that does not create more inflation, is something that they should actively do.

Sinclair: And so, in general, I think that this is where the economic debate lies, is what can the Fed do to create more jobs? Can they do something in a way that creates broad-based, more inclusive jobs or do we go back to our more simplistic macro models where it's just an aggregate number that they can create? I tend to lean, on this one, that it's pretty much just about aggregate, keeping the labor market hot, and that, in general, does look to spread out pretty well across the economy and across different types of jobs. And so, I think that maximum employment is as many jobs as we can get without causing inflation.

Beckworth: Well, I like that, and, I guess that I would maybe go one further and say that it makes me nervous if we did try to define broad-based and inclusive. I feel that we're getting beyond what we know, number one. Then, number two, it may be veering a little bit too much into fiscal policy. And I know that monetary policy and fiscal policy are not perfectly separated. They're always going to be linked, if nothing else, by the government budget constraint, but then, in practice, they are linked, no matter what we tell ourselves. So, I like your notion that we have to have some humility on what we know, but we have a goal for it.

Beckworth: So, the way you described it— I want to take that back to something that Milton Friedman said. I'm a big fan of Milton Friedman. In fact, there's a picture of Milton Friedman on the wall right outside my office, so if you ever come and visit my office, people will see this. But he was a big proponent of the plucking model which sounds very similar to what you've described, what this shortfall approach would be. Have you ever made the connection between those two?

Sinclair: I have, actually. Probably one of my very favorite papers I ever wrote— it does actually have plucking model in the title, I believe, in the final version. And so, I'm very much inspired by this idea that the economy does not look like a nice symmetric cycle over an upward sloping line. Instead, what we see is something more like an upward-sloping line, and most of the time, the economy is moving along that upward sloping line, but occasionally, it's plucked away, and that's exactly how Milton Friedman was describing the plucking model.

Sinclair: So, I did some unobserved components modeling of that type of model, and sure enough, I found that, basically— what I found, specifically, was that there's two kinds of recessions. One, where it is a Freidman pluck, where we just see this temporary pull away, but then other times we do see something where it really looks like that long run path changes as well, and so then we can also see a downturn from that as well. So, I think that there are two different kinds of recessions, more permanent ones and these transitory plucks. But in general, and in both cases, we do still have this sense of some concept of that maximum GDP growth, maximum employment. You can do it for different economic variables.

Beckworth: I guess, going back to the point you made, though, is the question of, well, what is it? What is the value, and we hit it? Because if it changes over time— And to be fair, every macro theory has this. I alluded to this earlier, the fiscal theory of the price level. They’ve got to know the real primary surpluses in the future, which we don't observe. Monetarists need to know real money demand. We don't see that. And whether you're looking at potential GDP or you're looking at maximum employment, whatever it is, you put it in your Phillips curve, and we don't see that either. So, I guess, whatever way we go, we have to be humble in what we do.

Beckworth: As someone who advocates nominal GDP level targeting, I guess the equivalent question would be, well, what's your trend path? What's a neutral nominal GDP level path? And in my mind, it's some measure of projected potential real GDP plus whatever your inflation needs to be. So, there's always going to be these measurement issues. I think that Hayek called it the knowledge problem. We simply don't know everything, so take a humble approach, but do your best, nonetheless.

Sinclair: Well, I think this brings us back again to the Romer and Romer presentation at the Brookings Papers conference, because they really highlighted that the forward-looking aspect is something that they want to see more of in the next framework review, and that makes me nervous, because I know how challenging it is to forecast the future, and in particular, when times are changing, and whether it be these transitory plucks, if you will— those are really hard to forecast. I have yet to identify a forecaster who consistently predicts them in the future without predicting a lot of false turns there. Then, similarly, there can be shifts around where that maximum employment number, potential GDP is. And so, that's where things get really tricky, because we want to be able to do forward-looking policy, and yet the Fed is throwing a whole lot of researchers at this forecasting problem and it doesn't really seem like it's a cracked problem at this point.

Could AI Define Maximum Employment in the Future?

Beckworth: Well, in the time we have left, I want to tie this discussion of the framework— which is a lot of fun and we could go on forever, probably, with it— and tie it back into our earlier discussion about data. I'm going to be very provocative here. Could we see a future where AI, big data, and whatever innovations we have in the future— it's able to divine, to figure out what maximum employment is in a way that we can't right now? So, maybe it's constantly, every day, sifting through the data like, “Oh, here's maximum employment.” Those parameters change, it figures it out. Is that a potential solution in the far future?

Sinclair: In other words, could AI come up with the proper specification of the Phillips curve?

Beckworth: Yes, or a Taylor rule. Yes, exactly. No, for sure, and it could update it in real-time and, of course, the implication that follows from that, which is one that Fed staffers don't want to hear, is that we don't need as many Fed staff. I think that you still need people to make the decision, because they're accountable. But there's a lot of jobs that would be affected, like you said. Maybe some go away, some new ones come. So, let's put aside Fed staffing decisions. That shouldn't drive our policy choice. But could you see a world where AI could better inform what maximum employment is?

Sinclair: I think that AI is already playing a role in potentially improving, or at least helping us to evaluate different forecasting models and frameworks. And so, I think that it already is a partner in some of this research, and so I'm very excited about that. And it's probably our best course forward at this point and trying to come up with really new ways of shaping the Phillips curve to get information from there. I'm very excited about that. I will highlight another direction, which actually would take us kind of back in time, but a new variation of a Friedman rule where we also have a much smaller Federal Reserve staff, but where we perhaps accept that we aren't able to predict the future and, instead, want to come up with something that is stable and consistent and [can] explain to the markets, might be another way that we look at going--

Beckworth: You know what? I have the answer to that. A nominal GDP level target, which we'll save for another time, another discussion. But with that, our time is up. Our guest today has been Tara Sinclair. Tara, thank you so much for coming on the program.

Sinclair: Thank you. It's my pleasure.