Perli: Reserve Management and the Fed’s System Open Market Account: Recent Experience and Insights from Surveys

Introduction

It is a pleasure to join this year's Financial Markets Conference. Thank you to the organizers at the Atlanta Fed for the invitation to speak today.

In my remarks this morning, I will first discuss recent money market conditions and the Open Market Trading Desk's (the Desk's) approach to reserve management purchases (RMPs). I will then discuss banks' demand for reserves based on results from the recent Senior Financial Officer Survey (SFOS), which the Federal Reserve administers twice a year. I'll conclude with some comments around our standing repo operations (SRPs), again using results from the SFOS as a starting point. Some of these considerations connect with the ongoing public debate about the future size of the SOMA portfolio and therefore about the size of the Federal Reserve's balance sheet.

Before I go further, I will give the usual disclaimer that these views are my own, and not necessarily those of the Federal Reserve Bank of New York or the Federal Reserve System.¹

Recent Money Market Conditions

I'll begin by reviewing money market conditions over the April tax season and recent decisions around RMPs.

As I have discussed previously, the Federal Open Market Committee's (FOMC's) decision to begin RMPs late last year was informed by expectations for a large and rapid drain of reserves in April amid flows into the Treasury General Account (TGA) during tax season.² Without RMPs, this would likely have pushed reserves below ample levels in April, potentially creating challenges for effective rate control.³

As expected, the TGA rose quickly to a peak of $1.04 trillion amid April tax receipts. While reserves declined by roughly $300 billion in the weeks surrounding April 15, prior RMPs ensured that reserves remained within the ample range. The April trough in reserves remained above reserve levels seen late last year, when the FOMC determined reserves had declined from abundant to ample (Panel 1).

Repo rates increased on the April 15 tax date, but the increase was modest and largely reversed the next day, with repo rates returning to levels near the interest rate on reserve balances (IORB) (Panel 2). As repo rates increased marginally above the SRP rate on April 15, some of our counterparties participated in SRP operations (Panel 3), which is exactly what I expected to see given that SRPs became economically sensible to use. I'll return to this point later.

All of this suggests that the RMP strategy initiated in December produced the desired results in terms of maintaining ample reserve conditions and ensuring interest rate control.

Since the tax date, we saw a decline in the TGA, a corresponding increase in reserves, and a softening in money market conditions, with the effective federal funds rate (EFFR) declining one basis point and repo reference rates printing notably below IORB in recent days. Negative net bill issuance has contributed to this softness. As a result, and consistent with previous communications, the Desk reduced the monthly pace of RMPs substantially and somewhat gradually, first from $40 billion to $25 billion for the mid-April to mid-May purchase period, and then to $10 billion for the mid-May to mid-June period (Panel 4).⁴

Looking ahead, the Desk will continue to manage SOMA securities holdings to keep an ample supply of reserves. I want to emphasize that RMPs are not on a pre-determined course. As I'll now discuss in more detail, we make RMP decisions on a month-to-month basis to preserve the flexibility that is needed to adapt to changing market conditions.

The Process for Determining RMP Sizes

The Desk's objective, as directed by the FOMC, is to maintain reserves within the ample range with RMPs. Our decisions on the size of RMPs are informed by three broad types of considerations, all equally important.

Our forecast for reserve supply, which is influenced by the demand for a number of Federal Reserve liabilities, is certainly a crucial part of the process. For example, as we saw in April during the tax payment period, the TGA can swing significantly, mechanically increasing or decreasing reserves in the system by comparable amounts but in the opposite direction; changes in other non-reserve liabilities have similar effects.⁵

We also consider how reserve demand is evolving, based on both our outreach and the analysis of survey-based measures of reserve demand collected through the SFOS. The recent survey covered nearly 100 banks that together hold 75 percent of the reserves in the banking system. It included questions on reserves and balance sheet management strategies as well as views on SRPs, which I'll discuss later.⁶

And last but not least, money market conditions are a key input and feedback mechanism. We consider signals from the various money market indicators that I've discussed in previous speeches, including the behavior of repo rates, their sensitivity to Treasury issuance, and their potential for exerting pressure on the federal funds rate.⁷ Signals also include usage of our SRP and overnight reverse repo (ON RRP) operations, changes in federal funds market dynamics, and banks' management of their payments and reserve positions.

In the months ahead, we will look carefully at all this information, and we stand ready to adjust the pace of RMPs up or down as necessary to maintain reserves within the ample range.

Some Survey-Based Thoughts on the Future of the SOMA

Currently, the SOMA portfolio stands at around $6.4 trillion, or 20 percent of GDP; as a result of the Committee's balance sheet reduction from 2022 to 2025, that number is well below the post-pandemic peak of about $8.5 trillion, or 33 percent of GDP (Panels 5 and 6). Assets are now slowly growing in dollar terms, via RMPs, to maintain ample reserves and accommodate growth in the demand for Federal Reserve liabilities. In principle, and everything else equal, assets as a share of GDP might remain fairly constant, though, assuming that demand for Federal Reserve liabilities grows more or less in line with GDP.

In real life, however, everything else is rarely equal. So, I'd like to spend a few minutes discussing possible factors that could change the size of the SOMA portfolio-not just in dollar terms, but also as a percentage of GDP. In doing so, I will build on results from the SFOS, which contained some interesting information about how respondents think about the issue.

Put simply, the size of the SOMA portfolio is a function of demand for Federal Reserve liabilities, the FOMC's choice of monetary policy implementation framework, and, within that framework, its tolerance for money market volatility.

The primary liabilities that drive the Federal Reserve's balance sheet size are currency (Federal Reserve notes), the TGA, and reserve balances; together, these components account for over 90 percent of current liabilities (Panel 7).

Currency has tended to increase with nominal economic growth and international demand for dollars; while potential growth in stablecoins and other forms of digital assets could change that in the future, to a first approximation, I expect currency to continue to grow at a more or less steady pace for at least several years.

The size of the TGA is managed by Treasury according to its cash management policy. Recently, the Treasury asked the Treasury Borrowing Advisory Committee (TBAC) for feedback on whether it should invest a portion of the TGA in overnight repo markets.⁸ At a high level, should the Treasury decide to do so, the TGA would decline by a corresponding amount; consequently, the Federal Reserve's assets needed to back the TGA would also decline by similar amounts or somewhat less, depending on implementation details. While the TBAC charge and other proposals have contemplated ways of reducing non-reserve liabilities, for the rest of my remarks I will focus on reserves.

The Federal Reserve implements monetary policy through an ample reserves framework, meaning that it supplies enough reserves to control short-term interest rates primarily through administered rates, rather than through the active management of the supply of reserves. "Ample" refers to the range of reserves that makes the federal funds rate only modestly sensitive to short-term variations in reserve supply (Panel 8).⁹

Let me make a brief comment on the performance of this framework. It has been in place de facto since the global financial crisis (GFC) and was affirmed by the FOMC in January 2019. A key criterion for any implementation framework is effective interest rate control across a range of economic and financial conditions, and on this measure our framework has an excellent track record: since December 2008, the EFFR has been outside of the target range on only two days.¹⁰ As I've discussed in the past, the ample reserves framework and the Fed's monetary policy implementation tools also support the resiliency of funding liquidity within the repo market, which in turn supports a well-functioning Treasury market and broader financial stability.¹¹

While the current implementation framework is demonstrably very effective, there is an active public debate about the quantity of reserve supply that it entails.¹² Conceptually, if policymakers wanted to reduce the supply of reserves and therefore the size of the Federal Reserve's balance sheet, they have two broad approaches at their disposal. The first is to take steps to reduce banks' demand for reserves and thereby reduce the quantities of reserves consistent with the ample range. We can think of this as shifting the reserve demand curve leftward, as illustrated in Panel 9. The second approach is to move up along the demand curve, as shown in Panel 10-that is, reduce reserve supply so that reserve conditions approach scarcity (or potentially even cross into scarcity, in which case the monetary policy implementation framework would change).¹³

A Shift in the Reserve Demand Curve

The current ample reserves implementation framework is well equipped to handle a reduction in the SOMA portfolio through a leftward shift in the reserve demand curve. At a high level, if reserve demand diminishes and reserve supply remains the same or increases via RMPs, there will be an imbalance between demand and supply of reserves, and that imbalance will be reflected in money market conditions just like it was when reserves were abundant.¹⁴ Depending on the specific source of the decline in reserve demand and the channels of transmission, it could be that the federal funds rate will soften, repo rates decline significantly, ON RRP usage increases, or that Federal Home Loan Bank advances will decline, or that a number of other market conditions will change. Regardless, some of the market indicators and survey data we collect and monitor will pick up the change; at that point, depending on the extent of the change, the Desk would be in a good position to adjust downward or stop the pace of RMPs or advise the Committee on the opportunity to resume balance sheet runoff. In this hypothetical but plausible situation, reserves would be lower than in the absence of a shift in demand, but rate control would stay strong and money market conditions would remain stable because reserve supply would still meet demand.

As I said, our current ample reserves framework would handle the task seamlessly. Of course, it is not the only possible efficient implementation of an ample reserves system-in fact, other central banks operate alternative versions while still achieving strong rate control. The Bank of England, for example, operates a system where the marginal quantity of reserves demanded by the financial system is supplied via temporary open market operations rather than via securities purchases.¹⁵ Such an alternative system, by itself, probably would not reduce the size of the SOMA portfolio by much; however, if policymakers desired to change the composition of the Federal Reserve's assets, it could work in the U.S. as well, with appropriate changes to the parameters and structure of Federal Reserve liquidity-providing operations.

There are many possible catalysts for a leftward shift in reserve demand, but a plausible one given the current debate is through potential future changes to bank regulatory liquidity requirements. Results from the recent SFOS support this idea.

According to the survey, the level of reserves that bank treasurers indicate they'd prefer to hold has been stable as a share of bank assets for the past few years, growing in nominal terms alongside the banking system. However, the survey also tells us that structural changes in the banking environment can meaningfully affect reserve demand. Banks cite changes to liquidity regulations, as well as shifts in liquidity management amid the transition toward 24/7 payments and the adoption of payment innovations, as important drivers of their preferred levels over the next two years (Panel 11).

These catalysts tend to push reserve demand in opposite directions, but survey respondents rated potential changes in liquidity regulations, which would reduce reserve demand, as the most important factor. This was especially the case for U.S. GSIB respondents and other large domestic banks that hold the largest quantity of reserves. This suggests that it is reasonable to expect a reduction in reserve demand if liquidity regulatory changes were to be implemented.

As a base case, surveyed banks generally do not expect a material change in reserve demand over the next year, likely because of the uncertainty as to whether regulatory changes will indeed be implemented, with no official proposals having been released to date. In addition, should changes materialize, respondents may think that it will take time to see them implemented, and possibly even more time for banks to react. Still, respondents indicated that risks are somewhat skewed toward lower reserve demand even over relatively short horizons.

Moving Along the Reserve Demand Curve

As I mentioned, reserves could also decline without a shift in the reserve demand curve-the FOMC could simply decide to supply a lower quantity of reserves and move the financial system along the existing curve. Under this potential approach, as reserve conditions tightened, money market rates would move persistently above IORB, and the opportunity costs of holding reserves would increase. Higher money market rates would likely be accompanied by higher volatility, which would introduce more risks in terms of rate control and funding market stability. Here too, the SFOS can shed some light on the likely response of money markets.

Panels 12 and 13 show aggregate responses to a SFOS question that asked banks what their level of reserve holdings would be given hypothetical increases in overnight rates. The brown lines correspond to the latest March 2026 survey and the blue lines to the previous survey from last September. The realized changes in bank reserve levels that we have seen between the two surveys have been broadly consistent with the same banks' responses to the September 2025 survey, so we can have some confidence in the reliability of their current projections.

Note that the curves are very steep, especially for domestic banks. This implies that banks would have to see large increases in overnight rates to be willing to shed modest amounts of reserves-or, conversely, that a modest decline in reserves could induce a substantial increase in overnight rates. This poses risks because we have learned from experience that increases in money market rates to levels even modestly above IORB can become disruptive very quickly, with attendant consequences not only for rate control but also for the stability of repo markets and, by extension, the Treasury market.¹⁶ Further steps to strengthen our SRPs would be useful in this hypothetical scenario.

Standing Repo Operations

I'll finish with some comments around our SRPs. These are a critical tool to ensure that the federal funds rate remains within its target range even on days of elevated pressures in money markets. As such, they are intended for the sole purpose of supporting monetary policy implementation, and they are expected to be used when economically sensible-that is, when market rates climb above the SRP rate.

Indeed, our primary dealer counterparties report that SRPs serve as an alternative funding source should private market repo rates exceed the SRP rate, thus supporting repo market intermediation at rates consistent with the FOMC's target range for the federal funds rate. The same can be said for our bank counterparties that engage in repo intermediation; in addition, banks report that SRPs offer a funding source that can be counted toward liquidity stress testing and used to address unexpected payment needs.

Past market outreach pointed to frictions that were discouraging use of SRPs by some counterparties and making the operations not as effective as they could be. Over the past year, the Desk and the FOMC have taken important steps to improve the effectiveness of SRPs; as a result, we have seen encouraging signs that willingness to use the operations has increased.¹⁷ For example, in February, March, and April this year, SRP operations were used when market repo rates were just barely above the SRP rate.

The encouraging recent experience is clearly backed by the SFOS and a related Desk survey of primary dealers.¹⁸ Bank counterparties reported that they are more likely to consider using SRPs than they were a year ago, when the same question was last asked. As shown on the left side of Panel 14, banks reported significantly lower "hurdle spreads"-that is, how far above the SRP rate market repo rates need to be for them to "actively consider" using SRP operations. Primary dealer responses (on the right) indicate even smaller hurdle spreads: the median dealer reported that they would actively consider submitting a proposition should private market tri-party repo rates reach around 10 basis points above the SRP rate, but several respondents reported much lower hurdle spreads.

Banks and dealers rated various factors affecting their views on SRP participation. The most positive factors, and also those that showed the most positive change since a year ago, were recent increased SRP take-up (which likely encouraged institutions that were previously reluctant to participate in the operations) and Federal Reserve communications (Panel 15). These improvements show that an effective Federal Reserve communication strategy and gradually increasing participation by our counterparties can materially shift the views of those who were previously reluctant to use the operations.

However, further steps to improve the operations may be helpful, especially if policymakers desire a smaller SOMA portfolio.¹⁹ SRPs provide certainty that liquidity would be available twice a day if there was a need, but only if they are perceived as efficient and counterparties are willing to participate in operations as intended. Under those conditions, SRPs can extend the lower end of the ample reserves range because they would dampen money market rate volatility. They could also help mitigate the risk I mentioned earlier associated with reducing reserves by moving along the reserve demand curve.

However, survey responses made clear there are still some impediments to SRP use, including disclosure requirements and the balance sheet costs of repo market intermediation funded by SRP usage. Centrally clearing SRPs would be one way to reduce such cost and thus could further improve counterparty participation and the efficacy of SRPs in curbing rate pressures and maintaining the stability of funding markets. Of note, some other central banks have also been moving toward central clearing to enhance the efficacy of some of their operations.²⁰ Centrally clearing SRPs would also align Desk operations with the broader U.S. repo market, which is increasingly moving to central clearing under the SEC mandate.^21,22

Still, a shift to centrally cleared repo transactions is a complex issue. Monetary policy implementation considerations would need to be weighed against other policy considerations, including those around central counterparty selection, membership models, and moral hazard.²³ Fed staff continue to evaluate this topic as well as other ways to potentially improve the efficacy of SRPs.

Conclusion

In summary, the Desk will continue to manage SOMA securities holdings to maintain an ample supply of reserves, but the monthly pace of RMPs is impossible to determine ex ante as it will be a function of varying market conditions. RMPs are aimed at meeting the demand for reserves and other Fed liabilities, which has been growing over time for a variety of reasons. If reserve demand declines exogenously, RMPs can be adjusted downwards or even stopped; in that case, the SOMA portfolio will shrink relative to GDP. If a decline in reserve demand is sufficiently pronounced, the SOMA portfolio can decline even in absolute terms. Either way, an ample reserves implementation framework is well equipped to handle these potential changes, especially if standing repo operations are used when economically sensible.

Thank you, and I look forward to the discussion.

Presentation

¹ I would like to thank Natalie Leonard, Eric LeSueur, and Linsey Molloy for their assistance in preparing these remarks, and Navya Sharma, Rachel Wilson, and Zack Youngblood for their assistance with the presentation.

² Roberto Perli, Reflections on the Early Days of Reserve Management Purchases and the Maintenance of Ample Reserves, remarks before the Money Marketeers of New York University, New York City, March 26, 2026. See also Roberto Perli, Eric LeSueur, and Linsey Molloy, The Implementation of Reserve Management Purchases to Maintain Ample Reserves, The Teller Window, March 31, 2026.

³ As I will discuss later, "ample" refers to that range of reserves that makes the federal funds rate only modestly sensitive to changes in reserves. See Perli (2026) and Julie Remache, Monetary Policy Implementation in an Ample Reserves Regime, remarks at the Fixed Income Analysts Society, Inc., Women in Fixed Income Conference, Federal Reserve Bank of New York, New York City, February 12, 2026.

⁴ See Federal Reserve Bank of New York, Statement Regarding Reserve Management Purchases Operations, December 10, 2025, and Perli (2026).

⁵ In the case of the TGA, consider what happens on a tax payment date: in general, as individuals pay taxes from bank accounts, the aggregate banking balance sheet contracts as deposits (a liability) and reserves (an asset) both decline by the amount of taxes paid. The size of the Fed's balance sheet doesn't change but the composition of liabilities does, with reserves declining and the TGA increasing. Any individual bank could choose to replenish its reserves, but those additional reserves typically come from other banks-that is, banks cannot change the total quantity of reserves in the system unless they borrow from, or temporarily sell securities to, the Federal Reserve.

⁶ Summary SFOS results are available at Board of Governors of the Federal Reserve System, Senior Financial Officer Survey.

⁷ See, for example, Perli (2026).

⁸ See U.S. Department of the Treasury, Minutes of the Meeting of the Treasury Borrowing Advisory Committee May 5, 2026, and Treasury Borrowing Advisory Committee, Treasury Investment in Repo, May 5, 2026.

⁹ Reserve supply that is less than ample results in scarce reserves, where the federal funds rate is highly sensitive to changes in reserve supply, and rate control requires active reserve management. Conversely, a reserve supply that is greater than ample results in abundant reserves, where the federal funds rate is unresponsive to changes in reserve supply and the quantity of reserves exceeds amounts needed for effective rate control.

¹⁰ The FOMC began using a 25-basis-point-wide target range for the federal funds rate in December 2008. The EFFR moved five basis points below the target range on year-end 2015 and five basis points above the target range on September 17, 2019.

¹¹ See for example Roberto Perli, Recent Developments in Treasury Market Liquidity and Funding Conditions, remarks at the 8th Short-Term Funding Markets Conference, Federal Reserve Board, Washington, DC, May 9, 2025.

¹² In recent months, there have been several papers and speeches related to the size of the Federal Reserve balance sheet, possible ways of reducing its size, and trade-offs involved with reducing its size. See Alyssa G. Anderson, Alessandro Barbarino, Anthony M. Diercks, Stephen Miran, A User's Guide to Reducing the Federal Reserve's Balance Sheet, Finance and Economics Discussion Series, March 2026; Michael S. Barr, Efficient and Effective Central Banking: Beyond the Balance Sheet, speech at the Money Marketeers of New York University, New York, New York, May 14, 2026; Darrell Duffie, The Payment System Puts a Floor on the Fed's Balance Sheet, Brookings Papers on Economic Activity, March 18, 2026; Burcu Duygan-Bump and R. Jay Kahn, The Central Bank Balance-Sheet Trilemma, FEDS Notes, January 14, 2026; and Lorie Logan and Sam Schulhofer-Wohl, Options for reducing the size of the Fed's balance sheet, April 2, 2026.

¹³ Some of the papers mentioned in the previous footnote contain ideas that fall under both of these general approaches.

¹⁴ As discussed in Perli (2026), reserves were abundant from mid-2020 to late 2025.

¹⁵ See Bank of England, Bank of England Market Operations Guide.

¹⁶ For example, EFFR was just a few basis points above IORB in the months leading up to September 17, 2019, when the repo market became clearly dislocated.

¹⁷ In June 2025, the Desk added a morning operation, with early settlement, alleviating an important friction for many counterparties. See Federal Reserve Bank of New York, Statement Regarding the Standing Repo Facility, May 28, 2025. In December 2025, former Chair Powell, at the FOMC press conference, publicly encouraged usage "when economically sensible", and the FOMC removed the aggregate operation limit. See Board of Governors of the Federal Reserve System, Transcript of Chair Powell's Press Conference, December 10, 2025, and Minutes of the Federal Open Market Committee, December 9-10, 2025.

¹⁸ See Dean Friedberg, Michael Koslow, Jason Miu, and Roberto Perli, Surveys Suggest Improved Willingness by Counterparties to Use Standing Repo Operations, The Teller Window, May 14, 2026. Additional information on the surveys is available on the New York Fed website here and here.

¹⁹ As President Williams has argued, the more frictionless lending operations are, the more effective they will be, and less of a reserves buffer will be needed to account for reserve demand uncertainty. See John C. Williams, On the Optimal Supply of Reserves, remarks at the New York Fed - Columbia SIPA Monetary Policy Implementation Workshop, Federal Reserve Bank of New York, New York City, May 22, 2025.

²⁰ See Bank of Canada, Bank of Canada to join the Canadian Collateral Management Service for its repo operations, March 6, 2026, and Eurex, European Central Bank to join Eurex Repo market, November 6, 2025.

²¹ See U.S. Securities and Exchange Commission, Treasury Clearing Implementation. See also Treasury Borrowing Advisory Committee, Central Clearing Implementation, May 2026.

²² An analogy could be the Desk's use of tri-party repo operations starting in 1999, which aligned the Desk with the evolving repo market structure and supported the efficacy and efficiency of Desk repo operations. See Federal Reserve Bank of New York, Domestic Open Market Operations During 1999. See also Antoine Martin and Susan McLaughlin, Financial Innovation: The Origins of the Tri-Party Repo Market, Liberty Street Economics, May 11, 2015.

²³ The minutes of the October 2025 FOMC meeting contain a summary of benefits and challenges associated with centrally cleared SRP operations as well as policymakers' views. See Board of Governors of the Federal Reserve System, Minutes of the Federal Open Market Committee, October 28-29, 2025.