Darrell Duffie on the Fed’s Balance Sheet and the Changing Financial System

Darrell Duffie is the Adams Distinguished Professor of Management, Emeritus, and Professor of Finance, Emeritus, at the Stanford Graduate School of Business. He joins the program to discuss the challenges of shrinking the Federal Reserve’s balance sheet, the future of payment-system design, the implications of tokenized finance, and much more.

Subscribe to David's Substack: Macroeconomic Policy Nexus

Read the full episode transcript:

This episode was recorded on September 18th, 2026

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: Welcome to Macro Musings, where each week we pull back the curtain and take a closer look at the most important macroeconomic issues of the past, present, and future. I am your host, David Beckworth, a senior research fellow with the Mercatus Center at George Mason University. I’m glad you decided to join us.

Our guest today is Darrell Duffie. Darrell is a finance professor at Stanford University and one of the leading authorities on financial market design and monetary policy implementation issues. He joins us today to discuss both, based upon some papers he’s recently done and presented at Brookings and at Jackson Hole. Darrell, welcome back to the program.

Darrell Duffie: David, it’s great to be back. Thanks for having me on.

Historical Context of Balance Sheet Discussions

Beckworth: It’s great to have you. It’s a real treat. Now, the last time you were on was way back in 2020. We talked about the Treasury market challenges of that time and reforms and a number of proposals that are now being implemented, like central clearing. Today we want to talk about the Fed’s balance sheet and what its future might look like.

As I mentioned in the intro, you have several wonderful papers on that. I want to maybe begin, before we jump into your two papers, by just setting the historical context of where we are. As you know, Darrell, in fact, you may be a part of these conversations behind the scenes, but we have a Fed Balance Sheet Task Force that’s happening right now that was initiated by the new chairman, Kevin Warsh.

We also have what seems to me, at least, is a change in openness among FOMC members, and I would say just those in the policy space writ large who care about this topic. There’s a few of us who’ve been beating this drum for a long time, but it just seems to me there’s more openness, more discussion. Maybe not everyone wants to go to a different operating system, but there seems to be an openness, at least, to shrink the liability side of the balance sheet. I’m wondering, how did we get here? How do you think about this moment that we’re in?

Duffie: I think it goes back even before the current chair or the speech by Kevin Warsh that he gave before he was chair at the IMF meetings last spring when he said that, in his view, the Fed’s balance sheet was too big. It goes back even further than that. The FOMC itself seemed to have been reluctant to have such a large balance sheet when it started normalizing in 2017, and the FOMC was very clear that it wanted the minimal balance sheet consistent with efficient monetary policy implementation.

It’s always been, I think, a point of sensitivity, although not a vividly sensitive topic at the Fed, that they didn’t like the look of a very large balance sheet, and that shows up in their transcripts. I think we’ve been coming this way, but with Kevin Warsh’s strong interest in reducing the Fed’s balance sheet, I don’t think it’s surprising that this is on the tip of everybody’s tongue today if you’re interested in central banking.

Beckworth: You recently had a paper out titled, “The Payment System Puts a Floor on the Fed’s Balance Sheet.” It was given at Brookings. We’re going to use that to motivate our discussion here in a minute. The timing of it was interesting because I think the same day you gave that, Stephen Miran gave a talk as well where he put out a paper, “A User’s Guide to Shrinking the Fed’s Balance Sheet.” Soon after that, Lorie Logan, Sam Schulhofer-Wohl gave out their paper.

It just seems like this crescendo is building. We’re reaching this point where people are talking about it. It’s great to see this moment. Again, I think it’s important that we have this conversation now because of this moment. One more question related to this moment, though. Do you think that the movement around the world in other advanced central banks toward demand-driven operating systems also is a tailwind for this conversation?

Duffie: Yes. I think this comes as part and parcel with what’s happening at the Bank of England, the European Central Bank, and the Bank of Japan even, which is reducing its appetite for JGBs. Central banks are generally reluctant to have such huge balance sheets, or in the past, were willing to do that. I think they’re going to be less willing in the future. In the meantime, they want to get smaller, but of course, I’m sure it’ll come up in our conversation that it’s not as easily done as said.

Beckworth: Your paper, again, is titled, “The Payment System Puts a Floor on the Fed’s Balance Sheet.” We’ll provide a link to it in the transcript. I’m sure many, if not most, of our listeners have already seen it. That’s who our audience is, Darrell. This is the perfect audience for you and your work here. Let me begin with a question to motivate this conversation on the balance sheet and shrinking it.

QT vs QE

Why is it that when the Fed does QT, it can’t fully unwind QE? I want to be very careful here. I don’t mean going back to the absolute level before it starts QE, but just that upward-moving trend path. We would expect it to be higher no matter what as the economy grows, but why is it so challenging for the Fed to return fully and offset the QE?

Duffie: I’d go first to a famous hypothesis presented at Jackson Hole in 2017 when Raghu Rajan presented his paper with Viral Acharya on a ratchet effect. The idea of their paper is that when the Fed blows up its balance sheet, it’s doing that by creating a lot more reserves. Those are the deposits that commercial banks have at the Fed. The commercial banks, they like those deposits. They’re really useful. They solve a lot of problems, including payment system issues, liquidity, they pay lots of interest, market interest rates.

Once they get used to those new supplies of reserves, they rejigger their operations so that they depend on having a lot of reserves. If the Fed tries to go back down in size of its balance sheet, eventually the banks find that they don’t have enough quickly available to meet their needs, and market interest rates get very spiky. The Fed tends to throw in the towel at that point and say, “Look, we got it as small as we could get it. We’re not going to get any smaller.” That happened again just recently in 2025 where the Fed’s balance sheet was starting to grow again when the Fed started to buy Treasury bills to keep the supply of reserves ample.

Beckworth: Do you think that’s a good way to also look at the 2019 repo moment where the rates soared during that time?

Duffie: Yes. That was the first major normalization. 2017 and 2018, the Fed brought down its balance sheet. Then, along the line of your question, it got as low as $1.4 trillion, which seems like a lot. It wasn’t enough at the time, but then more recently, the Fed was only able to get it down to about $3 trillion, which is more than twice as much. This ratchet effect described by Acharya and Rajan, it seems to be batting 1,000 in terms of its prediction that when the Fed increases its balance sheet, it can’t seem to get it back down as far as it was to begin with.

Beckworth: That is an interesting paper and finding, and like you said, it seems to be robust empirically. How much of that finding is an effect just going to happen no matter what else is happening versus an interaction effect because of all the liquidity regulations? We’ll get to those in a minute. But all the post-crisis liquidity regulations that were implemented, does that make it harder for banks to wind down their liability side, or is it something unique to QE and how banks respond behaviorally?

Duffie: I think there’s definitely an interaction effect with liquidity regulations, but if there is a ratchet effect, and it hasn’t been thoroughly tested yet, I think it could be that bank operations start to become based on the assumption of plentiful reserves. Once you’ve given your counterparties abundant lines of credit, allowed them to hold demand deposits in large volumes, and you no longer need to use interbank funding as much because you have tons of reserves, these systems atrophy, and your counterparties start to depend on these sources of liquidity, and then when you have to have less, it’s hard to reengineer your systems to do that.

Of course, the reason it interacts with liquidity regulations is that if you were to run out of reserves, you could, in principle, go to the Fed and get more. But that would demonstrate something counter to the original spirit of the liquidity regulations, which says you’re supposed to be self-sufficient, and being self-sufficient should, in principle, mean you don’t need to go to the Fed to get more. I think that’s when the tire meets the pavement, and you get these spikes in interest rates as banks start to become reluctant to lend money when they’re running out of enough to run their business operations.

Beckworth: Banks would, on their own, assuming no regulations are in play, would have this ratchet effect, but it’s compounded by the fact that they’re afraid or unwilling to go to the Fed to tap into the discount window or daylight overdrafts or the standing repo facility because of these regulations. Now, you mentioned a number of them. I’m going to throw them out there: liquidity coverage ratio, internal liquidity stress test, Regulation YY, CLAR, resolution planning. These are all in your paper. Which one of those you think are the most important? Are they all binding? Are they all creating this tendency to not want to use Fed facilities?

Duffie: For the purposes of our conversation, I think you can take the liquidity coverage ratio off the table. That’s not the one because that’s a 30-day test of sufficient liquidity. What we’re really seeing is evidence of intraday reserves balances being the binding constraint. When interest rates start to spike, and markets get very volatile, it’s because banks are holding back in the middle of the day on how much they’re lending.

The speed with which they’re paying each other and the other regulations that you mentioned, Reg YY, RLAP, the ILSTs, which are stress tests, and the CLAR, which is another test, those are all playing a role. They were all mentioned by Jamie Dimon famously after the September 2019 liquidity crunch when he was asked, “You had lots of reserves. Why didn’t you lend them early in the day when repo rates were skyrocketing? You could have made money.”

His reply on that earnings call was that the Fed is aware that there’s a red line at zero, basically, in terms of these liquidity regs, or at least JPMorgan had the sense at the time that if they were to go below zero, that would violate the spirit of these liquidity regulations, because it would demonstrate that they were not self-sufficient. Now, there’s nothing in the regs that say you shouldn’t overdraft. My impression is that the largest banks are starting to view these facilities or overdrafting as more available.

I think the Fed has been coaching them. We’re here for you. If you need liquidity, come to the Fed. That’s what these facilities are for, rather than viewing these regulations as preventing you from coming to the Fed. We may see some melting of that ice coming forward.

Beckworth: That’s interesting because I often will invoke the LCR. I’ve heard a number of officials invoke the LCR in these conversations, for example, about let’s make collateral at the discount window count toward the LCR. There’s a lot of emphasis on this. What you’re saying is, well, that may not be the most important one to look at. It may be more important to look at what these G-SIBs are facing, like the Regulation YY.

Duffie: Yes. I think that’s the case when you’re talking about the sufficiency of reserve balances in the system and what is minimally ample, depending on, in terms of liquidity regs, I’m quite confident it’s more about the intraday-based regulations.

Beckworth: Again, as a consequence of these regulations, these banks are less willing or not willing at all, in some cases, to go do daylight overdrafts, discount window usage, and standing repo operations, although that is a newer facility as well. Of course, discount window has some stigma, some other things associated with it. Your point, which I think is so true, is there’s not much use of it. Therefore, you lose your muscle memory. How to use these facilities, and when to use them? Let’s speak to another point that’s related to all this, and that is the supervisory expectations tied to this. It also compounds this problem. Is that right?

Duffie: Yes. It’s a question of perceptions. The regulations say one thing, and then the banks wonder, “Well, what is my supervisor going to say if I overdraft or if I go to the discount window or the standing repo operations?” Now, I have to say that we’re starting to see green shoots in terms of banks. The largest banks, or at least one of them, is, I think, now overdrafting in the middle of some days.

You saw, at the end of the year, $75 billion drawn from standing repo operations. You won’t see any action at the discount window by the largest banks because that’s still highly, highly stigmatized. I do think, if the Fed’s going to reduce its balance sheet, it’s going to be encouraging the largest banks to use overdrafts and standing repo operations more liberally. I think we’re going to start to see that. It’s still early days.

Beckworth: Again, just to reiterate this point, when we talk about these changes and reducing the liability side of the Fed’s balance sheet, we need to really look at the big banks more than the other smaller ones. Is that right?

Duffie: That’s true. The smaller banks are actually willing to overdraft, but it’s the largest banks that make markets for repurchase agreements, or repos. That’s the most important money market interest rate by far. It’s a $10 trillion-a-day market. It’s the largest dealer banks that make markets in repos. When they’re willing to lend funds overnight, at rates close to the Fed’s administered rate, which is called interest on reserves, then you know markets are fine. It’s the decision of those largest dealer banks that really matters.

Fedwire

Beckworth: We’ve talked about the Raghu Rajan ratchet effect. We talked about liquidity regulation. We talked about supervisory expectations. All of these are reasons why QT cannot completely unwind QE. There’s another story that you bring up in your paper. This is the central point of your paper. This is the inefficiency of the payment system with Fedwire. You propose a solution. Maybe walk us through that story and what your solution is for it.

Duffie: Sure. David, I’m pleased that you had a look at my paper. That’s one of the options that I put forward in my paper. Should the Fed want to reduce its balance sheet, one of the things that it could do is to add to its largest payment system, which is called Fedwire, a new piece of software which is known as a liquidity savings mechanism, or LSM. This is a piece of software that almost all large developed market central banks have, whether it’s the Bank of Canada, Bank of Japan, the Bank of England, the European Central Bank.

These central banks all have a piece of software that basically looks at the payments that the large banks are proposing to make and tries to match them against each other. So that if you are, for example, Bank of America, and you’ve indicated you want to make a $10 billion payment to me, you don’t necessarily have to take $10 billion out of your balances and make sure that you have those balances available because that $10 billion that you’re going to pay me might actually be offset by $10 billion I’m going to pay you, or $15 billion that you’re going to receive from Citibank. The liquidity savings mechanism is going to look for the opportunity to cancel your outgoing payments with your incoming payments. If it’s able to do that, then you don’t need to use reserve balances to make that payment. 

One of the options that I looked at in my Brookings paper is based on some research that I did with Chaojun Wang, a professor at the University of Pennsylvania, and Srisht Singh, who’s at the University of Toronto computer science department. We discovered a very simple way to match those outgoing and incoming payments in a way that banks would want to use. We show that it’s a dominant strategy for them to put their planned payments into the liquidity savings mechanism, get this cancellation, and avoid the use of so much reserve balances. That would allow the Fed to supply less reserves to the system and still have plenty of reserves for banks to use for other purposes.

Beckworth: To be clear, this is on the Fedwire system that you would apply this innovation?

Duffie: Yes. The Fed operates, as you probably know, a number of different large payment systems. The Fedwire system is the largest. It’s about $4.5 trillion of payments per day. Fedwire securities is also quite large. It’s over $2.5 trillion per day. Fedwire securities, which is the system by which you pay for Treasury securities with reserves, most of those payments, they’re not really subject to much discretion in terms of timing.

Most of them have to be made early in the day anyway, so banks are not holding back on those. It’s the Fedwire payments that are more discretionary and can be delayed significantly. When you delay your Fedwire payment to me because you’re worried you might not have enough, then I’m going to delay my Fedwire payment to another bank. Then that’s going to cause feedback because you’re going to be getting later payments from the other banks.

It can get into a vicious circle where we’re all waiting for each other. That’s called receipt-reactive payments, and it’s quite common when reserves are tight. As I showed in another paper with Adam Copeland at the New York Fed and Yilin Yang at the University of Minnesota, we showed that when reserves are tight, banks hold back on their payments to each other and market interest rates spike away from where the Fed wants them to be.

Beckworth: Can we tie this back into the liquidity regulations as well, that part of this reluctance to use a system efficiently is because they don’t want to be seen as not holding enough reserves, not holding that buffer, not being self-insured?

Duffie: Yes. It’s sometimes called a last taxi problem. They have these buffers, but if they use them, they might be violating the liquidity reg. A lot of this liquidity is trapped by regulation. Yes, that’s one of the problems. Liquidity regs may serve a useful purpose, but if they weren’t there, then banks would say, “Well, I don’t really care if I run short in the middle of the day. I can always go out and get some from the Fed, or I can overdraft.” Then you wouldn’t have some of these spikes that we’ve seen in 2019 and more recently in late 2025.

Beckworth: Yes, I was just wondering, and maybe you cover this in your paper, but Fedwire has been around for a long time, right? Did it fundamentally change with the advent of these liquidity regulations?

Duffie: Fedwire has been around since about 1980 in its current form. It was the world’s first real-time gross settlement system. It was a major achievement of the Fed at the time, leading all the other central banks. The issues with scarcity of reserves were actually more dominant even before the financial crisis. Yes, and banks did overdraft a lot in those days in order to make their payments. In those days—and you are one of the experts on this, David—the Fed wasn’t running monetary policy based on administered interest rates.

It was actually steering market rates by making reserves very scarce. In early 2007, for example, the whole system had about $10 billion of reserves. Today, we have about $3 trillion, so around 300 times as much. Back in those early days, banks were really having conniption fits some days trying to meet their needs to make payments and balancing that against the available supplies of reserves that they had. I’d say it’s better now, not worse, than it was pre-crisis in terms of stress on the system.

Beckworth: The only difference was back then they were more comfortable going to the Fed to make up for their shortage if they didn’t have it.

Duffie: Indeed. Not the discount window, but they didn’t mind overdrafting.

Beckworth: Yes. Since you mentioned the system, I’ll just throw this out there as an aside. If you go pre-2008, I don’t think we want to go back to that system, to be very clear. I don’t think we will go back, or we could go back because we didn’t have interest on reserves. We had what I would call an asymmetric corridor system. We did have an implicit interest on reserves, 0%. When fed funds rate goes to 5%, that’s a wide corridor. Banks would work overtime to get rid of reserves because the opportunity cost is so high.

You could imagine a world pre-2008 where you had a true corridor where maybe interest on reserves was, say, 50 basis points below the Fed funds rate. Banks would still try to economize on reserves, but they would hold more of them because there was some kind of compensation for them. We had a very unique system pre-2008; I don’t think anyone in this debate wants to go back because of that. Then, on top of that, what we’ve been talking about, liquidity regulations, they’re not going away. We can tweak them, and we don’t want to get rid of them altogether, but we could maybe make them more efficient. 

Duffie: Your last point, David. There’s a common perception that the Fed started to pay interest on reserves because of the financial crisis when it had to do quantitative easing and increase the quantity of reserves in the system. That’s actually not the case. It was the problems that you described with a scarce reserve system that caused the Fed to go to Congress much earlier than the financial crisis to get the authority to pay interest on reserves so that they wouldn’t have to deal with the scarce reserves regime and all of the stress that that put on both the Fed and the banks. Now, it happened that they accelerated the implementation of that with the financial crisis, but it wasn’t the original motive of the Fed.

Beckworth: That’s a great point. In my mind, completely understandable. It’s an implicit tax on the banks, and I would say an inordinate one. Again, when the fed funds raise 5%, and you’re earning 0% on your reserve, that’s a pretty big opportunity cost that the banks would face. You can see why the change was needed during that time. 

Going back to the Fedwire, just one other question here, maybe two more questions. The Fedwire is a real-time gross payment system. When you start to net—maybe that’s too strong of a term, but you’re netting—does that move you in the direction of a netting system? Because one of the challenges of a netting system is, as you know, it creates other demands upon you. The whole point of going to real-time is to get away from complete netting. What you’re suggesting isn’t a return or a movement to that, correct? It’s kind of a hybrid?

Duffie: It’s something in between. The old system, which is still used for ACH, which is one of the large payment systems, was based on what’s called deferred net settlement. With deferred net settlement, you would wait until the end of the payment period of the day, and then you would basically match all of your outgoing and incoming at that time, and you’d only have to pay the net, which involved a big reduction in the amount of reserves that you needed. That involves, unfortunately, credit risk. If you wait until the end of the day, maybe you won’t get paid some of the money you’re due, and you’ll be left holding the bag. 

Almost all central banks followed the Fed to RTGS, which avoided this credit risk associated with deferred net settlement. Over time, they found that indeed it required a lot of reserves because you have to stock up the reserves at the beginning of the day to make sure that you have enough and there’s no deferred settlement. Netting is not very effective in a pure RTGS system.

The other central banks, other than the US, figured out over time, “Well, we can economize and get some of the benefits of netting by running these liquidity savings mechanisms multiple times a day.” The Bank of Japan does it six times a day. That means there’s not much credit risk, but they get to cancel a lot of the outgoing payments against incoming, so the quantity of reserves that are needed is much smaller.

Beckworth: I love this about your work, Darrell. You look abroad, and you find these examples overseas, “Hey, let’s try this at home.” Of course, you have some models, to be clear. Your work’s very robust and rigorous, but you’ve looked overseas, and you see where other central banks are doing these things that we can learn from. I invoke that because I’m doing the same right now as I think about what could be an alternative operating system if the Fed were to go there. Now, I’m not sure the Fed will go there, and later in our program, I’ll bring up a few examples.

That was the first proposal, the main one, I think, in your paper, in terms of how you could shrink the Fed’s balance sheet, a reform that could happen. You bring up several others, so let’s walk through them. You also bring up the proposal to deal with unexpected shocks to reserves, the volatility in reserves from autonomous factors like the TGA, Treasury’s General Account, or we may be saying plain English, the federal government’s checking account at the Fed. Also, there’s the FIMA repo account, so foreign monetary authorities, they can come in and out of those. Walk us through how that’s important to shrinking the Fed’s balance sheet and what could be done about it.

Other Options to Shrink the Fed’s Balance Sheet

Duffie: Yes, good. I want to make sure that it’s clear. I’m not proposing that the Fed reduces its balance sheet. The purpose of my paper was to help with tools that could be used to do that if the Fed chooses to reduce its balance sheet. I think it’s very important that the Fed has those tools so that it doesn’t get backed into a political corner where it doesn’t have any way to respond to Congress or the administration with methods for reducing its balance sheet if told to do that, or if told what would be the consequences of not being able to reduce your balance sheet.

Maybe the Fed could be required, if it lost its independence, to buy assets that others might prefer. Again, these are not proposals per se, but they’re tools that I think the Fed should build and have available. Now, the one that you just asked about used to be used quite frequently by the Fed. It’s called temporary open market operations. I’m not the first to propose that. It was proposed by Bill Nelson some years ago, by Annette Vissing-Jørgensen at the Federal Reserve Board.

These TOMOs, or temporary open market operations, would be ways for the Fed to neutralize shocks to the supply of reserves so that they didn’t leave banks short on a given day. Think of it as a smooth road that you’re traveling down, and then there’s suddenly a pothole, and if somebody doesn’t come in ahead of you and fill in that pothole, your wheel is going to go into the hole and you’re going to have a big bump. These TOMOs are basically filling potholes that are created in a number of ways.

The simplest one is the fact that one of the biggest account holders at the Fed is the federal government. They have an account called the Treasury General Account, or TGA, and sometimes they pump up the size of that account, and the only way to do that is to take money out of reserves. The reserves supply goes down one for one with the TGA. Tax season is an example of that. When everyone pays their taxes, they’re taking money out of their banks.

The banks are taking money out of their reserves at the Fed and then paying the federal government into its TGA. When TGA goes up, you get a pothole in the path of reserve balances, and the Fed could, if it wanted to, temporarily fill that hole. It’s not been the policy of the Fed to do that for the last several years. Instead, the Fed wants a less activist policy on market operations.

The problem is that if the operations are less active, then the road is bumpier, and so you can either have smooth operations or a smooth road, but you can’t have both. Temporary open market operations, in the end, they wouldn’t make a huge difference in the quantity of reserve balances the system would need. It would be in the small hundreds of billions, if that, but it’s somewhat easier to do than adding a liquidity savings mechanism or some of the other things that I discuss in my paper.

Beckworth: It’s better to fill the potholes ahead of time than deal with the broken car down the road. Take care of things before they become a problem. You mentioned the Fed’s not doing them, and I wonder why. Why isn’t it? You said they’re reluctant to be activists, but where does that hesitation come from? Why are they so reluctant to actively go in and do it? Is it because it’s not part of the vision of the floor system, or what else could it be?

Duffie: This is a complicated question. One of the reasons is that the Fed has in the back of its mind what is called a demand-driven system for monetary policy implementation. What does that mean? Demand-driven means that when the banks discover a pothole, they will demand liquidity from the Fed. The Fed doesn’t have to come along in advance and fill those holes. If you rely on the banks to come to the Fed, then you’ll have a little bit more volatility of interest rates, and the reserves will appear because the banks will go to the Fed to get them.

Now the question is, will the banks actually do that? We discussed in the first part of this call the reluctance of the banks to demonstrate that they weren’t self-sufficient in terms of liquidity. Hopefully the banks will get over that, but until they do, this demand-driven aspiration of the Fed is only partially fulfilled.

In addition to this hope for a demand-driven approach, I think the Fed has a stylistic preference to have a less activist role in money markets. If you think of the machines that are filling in these potholes, scurrying up and down the road, finding the holes, filling them in, and scraping off the bumps, that’s busy work. I think some people at the Fed at least prefer to just set the level of the road and then see how it goes.

Beckworth: That’s interesting. The Fed would love for the banking system to become demand-driven in terms of the growth in reserves, but we have all these other obstacles we’ve just talked about, liquidity regulations, loss of muscle memory, that are preventing that from happening. At least the Fed’s heart is in the right place. It’s just there’s these obstacles to manifest them. That’s so interesting. That leads us maybe to our next conversation, which gets into some tweaks to the actual operating system.

Tweaks to the Operating System

You also mentioned the tiered reserve system, but I’m going to hold off on that a minute, Darrell, because I’m going to throw that out with some other possibilities. Of all the ones you’ve listed so far, I think you’ve answered this already, but just again, remind the listeners, how would you rank that? If you’re advising Chair Warsh, what is the easiest one, the next easiest, and go down the list?

Duffie: Yes, that’s terrific. In my paper, I do rank them, and it turns out that the ranking is also in order of how impactful they are on the quantity of reserves that the system needs. I would start with temporary open market operations being the relatively easiest to implement. Fed already has the authority. It’s used the tools in the past. It’s been supported, as I said, by Bill Nelson and Annette Vissing-Jørgensen. It’s pretty much a decision of the FOMC to go ahead and start doing that. Although it’s relatively easy, it’s not that impactful. At most, tens of billions or perhaps $100 billion or $200 billion, but certainly we’re not talking $3 trillion, not going to lop off the quantity that people are hoping to get rid of. It’s not in the trillions for sure.

Then if you go to liquidity regulations, those are somewhat harder to communicate in a way that is going to generate a demand-driven system. The Fed will have to do a lot of job owning. It’ll have to coach its supervisors, its supervisory staff, not to lean on banks that are drawing on reserves from the Fed. They’ll have to convince banks, “It’s okay, we’re not going to slap your fingers if you come to the Fed for reserves.” It can also, as it has done recently, clarify that their liquidity regulations don’t rule out going to the Fed for reserves. That’s number two on the list.

Number three on my list, which is more impactful, would be the liquidity savings mechanism. That’s likely to be in the hundreds of billions, although I wouldn’t want to put a number on it without doing more studies. I am continuing that research, and hopefully we’ll have some empirical work soon showing more of a prediction on how big that could get in terms of savings of reserve balances.

Then the last one, the biggest one, and perhaps the hardest to implement, mainly because of the communication difficulty and the pushback that the Fed might get, is the tiering of interest paid to banks on the reserve balances so that after they have enough to meet their business operations, the Fed would pay them at a lower interest rate. That would discourage banks from holding those reserves.

You mentioned earlier what bank wants to hold reserves that are earning an interest rate well below the market rate. They would try to eject those excess reserves, and that would allow the Fed to lower the quantity of reserve balances quite a lot. On the other hand, it’s hard to do because the banks are not going to like it, and it’s a complex rule to implement, although the Fed did use it at one time.

Beckworth: Yes. We’ll get back to this. I think it’s super fascinating. Going back to the liquidity regulations, since we didn’t actually spend a lot of time there in terms of the actual reforms of it, one of the ones that have been suggested a lot is to count collateral and other assets at the Fed’s discount window toward some of these internal liquidity stress tests, LCR. What would we do with Reg YY there? What would we do with that big one that you mentioned earlier to make it less onerous?

Duffie: It’s along the lines of what you just described. The regulation itself could actually stay as written. The regulation says that regardless of the circumstances that a bank finds itself in during a day, it must have the ability to meet its liquidity needs from its own resources. Now that’s the reg. What does its own resources mean? Does that include being able to draw liquidity from the Fed? Is that the bank’s resources, or is that going outside of the bank’s resources to get additional liquidity from the Fed? That’s a question of interpretation.

As you mentioned a moment ago, the Fed is dealing with what counts as the bank’s own resources when it’s drawing from the Fed. Right now, the Fed has clarified that if a bank has parked collateral, highly liquid assets, or HLA, at the Fed, the liquidity that it draws from the Fed against that collateral counts as the bank’s own resources in meeting Reg YY and the other regulations.

That’s a big plus if banks are convinced that that’s true because then they can draw on highly liquid assets. Bill Nelson, in one of his excellent papers on this topic, mooted the idea that the Fed could count not only highly liquid assets, but other assets that are parked at the Fed appropriately haircut for quality. If they were to be able to count those other assets toward their own resources, then it would be even easier for them to meet these liquidity regulations.

Beckworth: What’s the big “but” here, though? Is it that supervisors and bank management haven’t yet adapted to this new possibility?

Duffie: Two things. One is the Fed hasn’t done anything about collateral at the Fed that’s not highly liquid. It doesn’t let a bank count less than HLA. HLA is things like Treasuries. If it’s loans or other types of securities, they don’t count toward a bank’s own resources. The banks are still able to draw from the Fed based on those other forms of collateral, but they won’t count toward the bank’s own resources. The other thing that the Fed can do, other than clarifying collateral, is to coach the banks into the idea that if you draw on these resources, our supervisors are not going to bite you. 

Beckworth: That all takes time to change expectations and also to deal with the nonhigh-quality liquid assets. Let’s move on and talk about potential alternative operating systems. Now, I know in your paper, you didn’t list tiered reserves as an alternative operating system, but some might say that’s the case. I’m going to group that with some other proposals, but let’s do tiered reserves first since it’s in your paper.

I’ve done a lot of reading on at least the Norwegian example, and you mentioned that in your paper. The Swiss National Bank also does it, and some other countries have done it, but they’ve since left. Those are the two, I think, best examples currently. What I love about the Norges Bank, it actually had a floor system up until 2011. They’ve had experience with this for a number of years, long before we’re having this conversation today or even at the Fed a year or two back.

The Norges Bank in 2011, they actually go from a floor because they’re having many of the similar problems that we’re talking about now. If you read some of the speeches, and I’ve written this up in my Substack, but if you read the speeches from the officials at the Norges Bank, they’re worried, too, about all this liquidity hoarding. It’s becoming something they hold, not something they use.

Something else that’s really interesting—and you know this, Darrell—but they’re also really worried about the breakdown of the interbank market. Banks aren’t lending to each other, so reserves aren’t being efficiently distributed to where they’re needed within the banking system. They bring this system online. As you note in your paper, you’ve got a great chart that shows interbank lending. It picks up. It’s almost like this effect. If you look at total reserves, it had been growing under the floor system, and it comes down, and it’s been level ever since.

Now, to be clear, there are some other accounts at the central bank that are more of term deposits. It’s not quite as clean. It’s just looking at reserves because there’s other liabilities as well that the bank holds that’s tied to excess liquidity. Nonetheless, it seems to have worked relatively well, creating all the incentives that you’ve just outlined. You mentioned in your paper, and you mentioned a minute ago, that it’s radical. It would seem really radical. Some people have said, “I don’t like the Federal Reserve picking quotas. That’s very nonmarkety, very heavy-handed.” You suggest a voluntary reserve target. Let the banks pick the quota for themselves. How would that work?

Duffie: That voluntary reserves targeting was actually introduced by a committee within the Fed coming out of the financial crisis. When the Fed realized it was going to pay interest on reserves, it charged a secret committee. Later, its report was revealed publicly about seven or eight years later. This committee suggested one way to implement reserves tiering would be to have the banks report their own threshold at which they would get paid a lower interest rate.

You might say, “Why would a bank ever want to say anything other than my threshold is enormous because I want to get paid the high interest rate forever?” There are rewards and penalties for undershooting and overshooting that are in the voluntary reserves targeting regime. That’s been analyzed in a Fed paper by Garth Baughman, by the way, and Francesca Carapella. In their paper, they study the impact of VRT-based tiering.

I wouldn’t want to lend the impression that this is all theoretical. The Fed actually had a two-tiered system for a few weeks following the failure of Lehman. It was already in their monetary policy implementation. They had a strict separation between a high-tier rate for required reserves and then a low-tier rate for excess reserves. That was activated as soon as the Fed started to pay interest on reserves at the fall of Lehman.

Then, as we all know, the crisis turned out to be so severe that within a couple months, the Fed dropped its interest rates to the zero lower bound, and so both the upper tier and the lower tier got squashed to zero. That’s effectively a one-tier system. When eventually the Fed lifted its administered rates above zero, it did them together. We’ve never gone back to two tiers.

Again, I just want to emphasize this is not only a theoretical possibility. The Fed did this for a short period of time. It could do it again. It would be radical mainly because of two reasons. One is banks are going to say, “What the heck? What are you doing to us? You’re not going to pay us the interest rate on all of our reserves, only on some of them? We don’t like that.” They might complain about that, including to their congressional representatives.

The second concern is the one you mentioned. How do you know how much reserves a bank needs, which would be the basis for the high interest rate, and how much is excess, which would be the basis for the low interest rate? That’s hard to determine. If you do it liberally, you’re still going to save a lot of reserve balances for the system.

You mentioned earlier the Norges Bank, the central bank of Norway, being one of the early implementers of this approach. Even earlier than that, the Reserve Bank of New Zealand, which is a leader among central banks in new monetary policy operational approaches, implemented a two-tiered system very successfully in 2007 with the express purpose of reducing the size of its balance sheet. It said explicitly, “We don’t want banks holding reserves as an investment. We want them to have enough to meet their payment needs, but not more. If we tier, then they won’t want to hold the extra balances, and we can reduce the size of our balance sheet.”

Beckworth: Nothing new under the sun. These are questions that other central banks have looked at before. The floor system isn’t something unique to the US. Other central banks have dealt with the challenges and have adapted. Tiered reserves seems to be, at least among several central banks, a tried and true way to deal with this. Again, I want to stress the interbank angle on this. If you go to a tiered reserve system, what happens is some banks may be short one-day reserves, some are long, and so they start lending to each other. To me, a markets guy, I think that’d be a wonderful development. There’s price discovery. There’s counterparty discipline.

I’m not saying you bring back the federal funds market like its glory days, but you do something that there’s extra price signals, extra price discovery. I would note that it’s not just David Beckworth, Bill Nelson, who’s making his point. Many of the central banks around the world that are going to demand-driven systems, they explicitly state they want to see interbank lending resurrected again.

We’ll go to that, but maybe, Darrell, speak to that. Do you think that is a goal worthy in itself? I have brought this up with other guests, and I’ve read other accounts, and they’re underwhelmed and say, “So what? We got repo markets. We got stocks on banks. Why do we need another market for banks?” How would you respond to that? Maybe more generally, what’s the point of reviving the interbank market?

Duffie: Consistent with your remark, David, we now have a federal funds market that is almost inactive among US domestic banks. It’s on the order of $5 billion a day between US domestic banks borrowing and lending from each other in the federal funds market. The repo market, as I mentioned, is on the order of $8 to $10 trillion per day. Forget about fed funds. Why don’t we just use the repo market? Not all banks want to use the repo market. It requires setup costs. It’s mostly a large bank phenomenon. There’s a lot of medium and smaller banks in the US that would use the federal funds market if they had enough incentive to go to it and if the frictions were smaller.

In some research that’s ongoing with two Stanford PhD students, Thanawat Sornwanee and Francesco Spizzuoco, we show that, in fact, the difference between the upper-tier interest rate and the lower-tier interest rate needs to be big enough to overcome the reluctance to trade due to frictional costs of accessing the interbank funds market. Yes, reinvigorating that market would probably reduce some of the ratchet effect that we spoke about earlier in this call, by which banks maybe have lost their facility with accessing the federal funds market because over the last 10, 15 years, they’ve had tons of reserves. They haven’t needed it.

Then when it comes time that they need it, they don’t have a well-oiled-up operation. Access and reinvigorating the federal funds market would, in my view, be a good idea. Maybe I have a different set of friends than you do, but everyone that I’ve talked to said, “I’m not sure how important it is, but yes, it would be a nice thing to have.”

Beckworth: Extra price discovery, extra signals—it’s useful. Again, better allocation of reserves in the banking system, I think, is a key point. 

I’m going to have to jump to your second paper due to time, but I’ll just mention what I was going to ask you, Darrell is we have, again, these demand-driven systems in Europe. I think the Reserve Bank of Australia is probably on the cutting edge of it. I will direct listeners and watchers to my Substack. I have an article there. I find it really intriguing what the possibilities are there. It would take some changes at the Fed for that as well. Then, as you know, Darrell, because I’ve sent you my stuff, I’m really enamored with the Riksbank. They target a term deposit facility rate, which I think might also deal with some of the challenges of the ratcheting effect. I’ll direct listeners and watchers there. 

Tokenized Finance

I want to get to your awesome paper that you just delivered at Jackson Hole because, again, it has a bearing on the question about the size of its balance sheet and what the Fed can do. Its title is “Tokenized Finance and Perimeter of Central Banking.” Maybe give us the quick executive summary of it, and then we can delve into the details.

Duffie: I think the executive summary is that it’s not about Bitcoin; it’s about the backend of the financial system where Treasury repos and large foreign exchange trades are done, security settlement. That’s going tokenized. You can’t say exactly how many years, but the largest banks, the DTCC, international central banks, they’re all sharpening their knives to prepare for the ability to trade traditional financial securities on programmable ledgers.

Now, they might not be exactly blockchains, but they’re going to have the features that we think about on a blockchain-based system where the assets can be exchanged using smart contracts, which are forms of software that allow a lot more programmability of your trades as to when you trade, under what conditions you trade, the ability to trade without requiring a central counterparty for every trade, and so on.

The world is going that way sooner or later. My paper basically describes why is that happening, and are central banks prepared for it? The answer is no, they’re not prepared yet, although some central banks are moving quickly in that direction. The reason that the ground is not properly prepared for tokenization is that once that movement gets big enough, there’s no safe enough cash settlement asset available to programmable ledgers today.

You might say, “Why not stablecoins? That’s a programmable form of cash,” but there are no stablecoins safe enough to settle multitrillion-dollar-a-day markets. Any stablecoin that I know about has a price that fluctuates close to $1, but it’s not exactly $1. If we do a billion-dollar repo, and I assure you that the cash I’m sending you in the form of stablecoins will be worth at least $0.999, you’re going to say, “What? I’m not giving up $0.001 of a billion dollars. No way. I’ll lose my job.”

The unit of account for the largest financial markets must be central bank money or equivalent. There’s no question about what the price in dollars is. It should be the dollar. That was long ago agreed, even before the advent of tokenization, that large-scale financial market infrastructure should be settled only in central bank money. You don’t want to have financial stability resting on whether commercial bank deposits or some other form of money is safe enough because otherwise, on a day on which it might not be, you could get a sudden stop and the largest financial markets could come to a halt and cause a financial crisis.

Beckworth: Tokenized finance has lots of potential, lots to offer, 24/7 trading settlement. You talk about atomic delivery versus payment in there. Lots of potential. Currently, it doesn’t have the payment infrastructure to handle large-scale transactions. You talk about there’s this potential for fragmentation, and that could increase the demand for the Fed’s balance sheets. Maybe talk about the fragmentation concern and the demands it would put on the Fed.

Duffie: We’ve just spent the previous half hour, David, talking about the fact that central banks are interested in having smaller balance sheets. If tokenization starts to get big, which it eventually will, then you’re going to have a bifurcation between conventional money and programmable money. Same thing with securities. You’ll have conventional securities and tokenized securities.

Every large financial market participant will have to have two forms of everything, and they won’t be immediately fungible with each other, and so the quantity of money and securities you need to start each day will be larger. Your funding costs will go up. Central bank will try to mitigate that by supplying more reserves to meet the greater need for cash. Central bank balance sheets will, at least initially, probably get larger. That’s the cost of fragmenting between conventional and new forms of payment and settlement systems.

Even within the new forms, it’s not settled yet what kind of ledger. We’re going to have multiple different ledgers on which this is all happening. A leading candidate right now is called Kenton, but other programmable ledgers could come along. If they don’t talk to each other on a high-frequency basis, that will also increase fragmentation costs, as everybody will have to preload these different ledgers, each with the amount of money that they need for the day.

Duffie’s Recommendations

Beckworth: What are your recommendations? What’s the ideal path forward to solving this dilemma?

Duffie: First, the central banks will have to accommodate. They can’t be stingy with creating enough reserves to serve the increased needs for reserves, because otherwise, we get back to the conversation we had a little while ago, market interest rates will spike when there’s not enough reserves in the system. Central banks will do that. That’s the first order of the day. To make the system safe, as I said, they would need to provide reserves in a tokenized form in addition to conventional form, or figure out how to make their existing ledgers programmable so that they can do both with one pool of cash.

The next order of the day is really up to the private sector. How are they going to make better use of these new forms of payment and settlement to do more frequent funding and more frequent settlement so that they don’t need as much cash? That’s a different dimension, which is the frequency with which you replenish your accounts. In today’s conventional markets, funding is a one-day, or more, transaction.

Overnight repo, that huge market we discussed, that’s one-day settlement. In the new world of programmable finance, you will be able to do funding cycles multiple times a day. We already see Broadridge doing on the order of $8 trillion a month of intraday repo contracts that settle within the day, not on the next day, allowing banks to manage with less liquidity at the start of each day.

Beckworth: You mentioned the central bank, the Fed itself, using a programmable ledger, maybe even smart contracts. Circling back to our earlier conversation, Darrell, when we think about could central banks have a demand-driven operating system, can you imagine the discount window, daylight overdrafts, standing repo operations, those being done on a programmable ledger with smart contracts? It’s basically liquidity on demand or just-in-time liquidity. You don’t even need to worry about logistics. It just happens automatically. Therefore, you could offset some of this need for extra liquidity, pre-funded liquidity in the system.

Duffie: That exact question came up from Don Kohn during the Jackson Hole. Just after my Jackson Hole presentation, he asked a question, how will central banks meet this high-frequency, 24/7 programmable financial market when they provide liquidity to the markets? I reacted as you just did. They will have to provide them in a way that’s accessible, at least to programmable ledgers, and probably overnight and on weekends as well. They’re already moving in that direction.

By the way, I want to clarify, I’m not expecting the Fed to be one of the early movers in this among central banks. The ECB is moving hard on providing tokenized central bank deposits to its financial system. The Swiss National Bank has already done that. The Bank of England is figuring out how to do this with conventional reserves that are available on programmable ledgers. Other central banks are moving ahead: Australia, you mentioned the RBA is moving ahead in this area.

The Fed, at least, in terms of what we’ve seen at this point, I think it’s watching and waiting to see what technology should we use, should we let the private sector lead on this, should we supply tokenized reserves, or should we allow special purpose firms to provide tokenized deposits backed by central bank deposits? That’s one step removed but still very safe. That’s a distinct possibility in the United States. I don’t expect the Fed to be an early mover on providing tokenized reserves.

Beckworth: That’s probably a good thing. We’re the big ship that slowly moves. Bank of New Zealand, maybe the Swiss National Bank, they’re small, nimble. They can turn quickly and experiment, and we can learn lessons from them. It’s still a pretty exciting vision of the future. If the Fed had a programmable ledger, smart contracts with its liquidity facilities, I think that’d be an amazing world. Something to look forward to. 

Darrell, thank you so much for your time today. Again, his papers we’ve discussed, you need to check out: “The Payment System Puts a Floor Under the Fed’s Balance Sheet” and “Tokenized Finance in the Perimeter of Central Banking.” Darrell, thank you once again for coming on the podcast.

Duffie: David, it’s such a pleasure to be with you, and thank you for inviting me.

Beckworth: Macro Musings is produced by the Mercatus Center at George Mason University. Dive deeper into our research at mercatus.org/monetarypolicy. You can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. If you like this podcast, please consider giving us a rating and leaving a review. This helps other thoughtful people like you find the show. Find me on Twitter @DavidBeckworth, and follow the show @Macro_Musings.