Efficient and effective central bank balance sheets

Thank you, Vicky [Saporta], for the kind introduction. And thank you to the Bank of England for inviting me to participate in this important conference.

The experience of the past two decades profoundly changed how central banks think about both sides of our balance sheets. On the asset side, central banks learned to use asset purchases to provide economic stimulus at the effective lower bound on short-term interest rates, as well as to support smooth market functioning when stresses threatened financial stability. On the opposite side of the ledger, asset purchases drove large increases in central bank liabilities, specifically bank reserves, making liquidity abundant or ample rather than scarce in most financial systems.

Experience has shown that ample liquidity provides ample benefits for financial stability, payments and monetary policy implementation. Central banks are now pursuing strategies to maintain those benefits even as the runoff of pandemic-era assets reduces the size of their balance sheets.

At times, the diversity of names and tools for these strategies can make it seem as if central banks disagree on the fundamentals of how to proceed. People talk about demand-driven and supply-driven floors, ample and abundant reserves, zero corridors and wider corridors and more. But underneath this cacophony of details, I believe central banks are applying shared and well-founded principles for how their balance sheets should work. The surface differences, to my mind, reflect central banks’ natural adaptation to unique characteristics of the financial systems in which they operate.

In the rest of my remarks, I’ll lay out the shared principles as I see them and highlight their implications for the Federal Reserve’s balance sheet. I’ll focus primarily on liabilities before turning to assets at the end. As always, these views are mine and not necessarily those of my Fed colleagues.

Two principles

I propose that in employing our balance sheets to accomplish our missions, central banks should adhere to two fundamental principles. We should be efficient. And we should be effective.

You may recall hearing those ideas from the Fed before. Back in 2014, long before adopting our current implementation framework, the FOMC said it aimed to “hold no more securities than necessary to implement monetary policy efficiently and effectively.”[1] We’ve repeated that idea in subsequent statements.[2] Today, I’ll examine efficiency and effectiveness as broad principles for central banks, beyond the context of any one Fed policy decision.

To some, the principles I’ve offered may sound so obvious as to mean nothing. Who could want a central bank to be inefficient or ineffective? But I’ll argue that these principles have a great deal of meaning. They provide substantial guidance about the design of central bank balance sheets, and they create a clear framework for understanding central banks’ choices. I’ll consider each principle in turn.

Efficiency

By efficiency, I mean allocative or Pareto efficiency. The central bank should design its balance sheet in such a way that no adjustment could make someone better off (in the economy the central bank serves) without simultaneously making someone else worse off (again, in the economy the central bank serves).

Allocative efficiency requires setting the price of resources equal to their social marginal cost. Let’s apply that idea to central bank liabilities, specifically, reserves. From the perspective of a commercial bank considering how many reserves to hold, the price of reserves is the spread between money market interest rates and whatever rate the central bank pays on reserves. That is, when money market rates exceed interest on reserves, the commercial bank pays an opportunity cost in foregone interest to hold reserves.

The central bank’s marginal cost of supplying reserves is typically small. Reserves are an electronic entry on the central bank’s books. No paper currency needs to be produced.

So, as I have argued previously, efficiency calls for money market rates to land close to the interest rate on reserves.[3] If money market rates exceed interest on reserves, banks experience inefficient pressure to economize on reserves, the most liquid asset in the financial system. That increases financial stability risks and can gum up the flow of payments. The opposite problem occurs when interest on reserves exceeds market rates. Typically, only a limited set of financial institutions, such as banks, can hold reserves. Other financial institutions must hold their liquidity in other instruments, such as repurchase agreements (repos) or short-dated government securities. When these instruments pay less interest than reserves, the cost of liquidity is inefficiently higher for non-banks.

Elsewhere, I’ve described the idea of keeping money market rates near interest on reserves as a Friedman rule for interest-bearing reserves.[4] There’s more than one way to achieve this configuration of rates. A central bank can estimate the quantity of reserves that would meet banks’ demand with market rates near interest on reserves, then acquire sufficient assets to supply that quantity. Or, like the Bank of England, a central bank can offer to lend reserves at the same interest rate it pays on its deposit facility—letting banks take down however many reserves they demand. Likewise, if an excessive supply of reserves pushes money market rates below interest on reserves, a central bank can drain liquidity and bring rates back up by reducing its asset holdings. Or, as with the Fed’s Overnight Reverse Repurchase Agreement Facility, a central bank can offer to absorb liquidity from market participants ineligible for interest on reserves.

The optimal choices among these tactics may depend on the specifics of a central bank’s environment, such as its counterparties, how liquidity flows between those counterparties and other firms, the counterparties’ willingness to borrow from central bank facilities, the central bank’s institutional authorities, and the exogenous influences on reserve supply and demand.

But, again, those choices are tactical. The strategic decision is where to position interest on reserves relative to money market rates. And when I look across major central banks, it seems to me our approaches are fundamentally similar. We all aim to keep money market rates close to interest on reserves, which is the efficient arrangement.

You may have noticed that I said efficiency calls for money market rates to settle close to interest on reserves, not necessarily equal to interest on reserves. Technical factors can appropriately create small spreads between different money market instruments. For example, the repo market in the U.S. has several segments.[5] Small rate spreads between segments can reflect participants’ market power or transactions’ non-price terms. No amount of reserve supply can compress these spreads to zero, nor could the interest rate on reserves match all the different repo rates at once.

Similarly, even in an efficient configuration, temporary frictions such as those often seen on statement dates could cause short-lived rate movements. Such fluctuations aren’t necessarily inefficient. Suppressing them could require supplying so many reserves as to push money market rates materially below interest on reserves the rest of the year.

The cost of supplying reserves also may not be exactly zero. Yesterday’s keynote speaker, Annette Vissing-Jorgensen, has pointed out that if a central bank backs its liabilities with assets that carry a scarcity premium, removing those assets from circulation is costly.[6] I’d note, though, and I’ll return to this later, that the cost depends on whether the supply of such assets can grow.

Some observers have objected to keeping money market rates close to interest on reserves, saying this suppresses trading in the interbank money market. In the face of unexpected payment shocks, an active interbank market helps the private sector redistribute funds across banks. By comparison, without an interbank market, the central bank may intermediate more of that redistribution through its deposit and lending facilities. Policy preferences can differ about the desirability of trading off a smaller central bank footprint against the spread of market rates over interest on reserves, and the resulting inefficiency, that would be needed to incentivize active interbank trading.

While that’s an interesting philosophical discussion, I don’t see much of a practical argument for the U.S., at least. When the U.S. banking system has excess reserves, they tend to accumulate at the largest money-center banks. The regulations applicable to such banks disincentivize unsecured lending to other banks. As a result, only very large spreads between money market rates and interest on reserves would likely suffice to revive the U.S. interbank market—and that would be very inefficient.[7]

Nor would an active interbank market improve banks’ resilience to liquidity stress. Making liquidity artificially scarce and then giving banks lots of practice managing that scarcity is riskier and less efficient than avoiding scarcity in the first place. And there are plenty of other ways that banks can maintain operational readiness to exercise their liquidity tools.

Effectiveness

I’ll now turn to my second principle, effectiveness. A central bank’s tools must reliably transmit the stance of policy to money market rates and the economy. Researchers have debated how exactly to define effectiveness.[8] Is it about the dispersion of rates within or across money markets? Volatility over time? Robustness under stress? Transmission to broader markets? The answer to that technical question may differ across financial systems. For purposes of principles, it suffices to say that effective implementation keeps money market rates close to the policy target.

A central bank can influence market rates through its balance sheet in two main ways. One approach supplies a limited quantity of reserves, so that reserves carry a liquidity premium and market rates significantly exceed interest on reserves. The banking system then typically lands on a steep portion of the aggregate demand curve for reserves. The central bank achieves its policy target by increasing or decreasing reserve supply to intersect the appropriate point on the demand curve. In the alternative approach, the central bank makes the supply of reserves ample enough to eliminate the liquidity premium—the efficient configuration I described with my first principle. The banking system then typically lands on a relatively flat portion of the aggregate demand curve for reserves. Small changes in reserves supply don’t substantially move money market rates in that environment. Instead, the central bank moves money market rates primarily by adjusting the interest rate on liabilities such as reserves.

Before the Global Financial Crisis (GFC), scarce-reserves systems provided effective rate control for the Fed and many other central banks. And systems with ample reserves have proven effective in controlling rates more recently.

While it remains conceptually possible to implement monetary policy today with scarce reserves, developments since the GFC would complicate the task of fine-tuning reserve supply to hit a point on the steep portion of the demand curve. Changes in regulations and banks’ own risk management have made reserve demand larger, more volatile and more difficult to predict. In the U.S., fiscal flows also create more reserve volatility than previously.

The ample-reserves regime simplifies rate control because fluctuations in reserve supply and demand don’t require frequent and precise offsetting central bank actions. You might say this makes the ample-reserves regime efficient for a second reason. Not only does it avoid the inefficiencies associated with a liquidity premium on reserves, but it also saves some forecasting and operational effort. The ample regime requires gradual forecasting and operations to approximately track demand for central bank liabilities over periods of months and years. The scarce-reserves regime requires the much heavier lift of actively managing reserve supply on a daily basis to precisely match demand.

Ample-reserves regimes are sometimes called “floor systems,” because when reserve supply is large enough, the main influence on market rates is the floor established under them by interest on reserves.

But the “floor system” name can give a misleading impression of how these regimes work. Even in a floor system, central bank lending tools — which put a ceiling on interest rates—retain an important role in policy implementation.

Central banks’ estimates of the quantity of reserves needed to remain at the floor, though careful, are less than perfectly precise. As a result, unexpected shortfalls of reserves can sometimes generate upward pressure on money market rates. Such events arise much less frequently with ample reserves than scarce reserves—that’s what makes the ample-reserves approach more effective—but the incidence isn’t zero. So even in an ample-reserves regime, reliable ceiling tools remain indispensable, at minimum as a backstop against mistaken estimates, and potentially as a routine source of reserve supply. Indeed, many central banks plan to routinely employ one or more ceiling tools in their long-run implementation frameworks. While these tools’ designs vary in response to local needs, the common principle is that they make policy implementation more effective.

The Fed’s ceiling tools include the discount window, the Standing Repo Facility (SRF), and the Foreign and International Monetary Authorities Repo Facility, as well as the Open Market Trading Desk’s capacity to conduct discretionary repo operations.

Although these tools are time tested and have often worked well, I believe we could further enhance their efficacy.

The discount window is flexible and robust, capable of lending to banks against a very broad range of collateral throughout the day. For this reason, I believe every bank in the United States should make sure it is operationally ready to access the window. Banks have increased their operational readiness since the 2023 stresses, but readiness should be a continuing effort, not a one-time project. And it needs to be a partnership between borrowers and lenders. At the Federal Reserve Banks, we are considering how we can most efficiently serve our customers when they come to borrow. The Board of Governors recently completed a Request for Information on discount window operations.[9] I’m optimistic that feedback will help us support enhanced readiness, particularly by considering more automation, extended hours of operation, faster communications with depository institutions and interoperability with other secured lending sources.

Turning to the SRF and other open market operations that provide liquidity, the Fed’s tools will function most effectively if they innovate alongside the markets in which we participate. The Desk experimented with morning SRF operations over year end and has announced an upcoming small-value test of capabilities for morning settlements.[10] And as the U.S. repo market continues to shift toward central clearing, I believe the FOMC should consider the merits of centrally clearing the SRF as well as other open market operations. Centrally clearing the Fed’s trades would support the Securities and Exchange Commission clearing mandate for the private sector. It would also decrease our counterparties’ cost of intermediating funding to the broader market.

Trade-offs between efficiency and effectiveness

I’ve argued that, at a high level, supplying ample reserves and positioning money market rates close to interest on reserves is both efficient and effective. Still, when it comes to the details of implementing such a regime, central banks can face modest trade-offs between efficiency and effectiveness.

These trade-offs could fill an entire speech, or a book. Today, I’d like to discuss just one of them, the precise positioning of money market rates relative to interest on reserves.

In the U.S., the volatility of money market rates picks up noticeably as market rates move above interest on reserves.[11] The increase in volatility appears to stem largely from frictions in redistributing liquidity among financial institutions. In this environment, operating with money market rates close to, but perhaps slightly below, interest on reserves is likely to provide more effective rate control. In my view, for the U.S. financial system, bringing market rates all the way up to interest on reserves could reduce effectiveness while delivering only a minimal improvement in efficiency.[12]

Some other central banks are planning for money market rates to converge slightly above interest on reserves. Those central banks operate in different institutional environments and face different frictions, which can change the trade-off between efficiency and effectiveness and may support a different configuration of rates than in the U.S.

Assets

So far, I have emphasized the central bank’s liabilities. Efficiency and effectiveness also apply to the assets backing those liabilities.

Central banks vary greatly in the assets they can hold and their relationships to fiscal authorities. Hence, asset choices could differ more than choices about liabilities. I’ll focus my discussion on the U.S. case, but the principles I’ll apply have broad relevance.

The FOMC has said it intends to hold primarily Treasury securities in the long run.[13] This approach leaves open two questions: Are some Treasury securities more efficient or effective to hold than others? And, against the backdrop of a primarily Treasury portfolio, might efficiency or effectiveness call for holding modest amounts of other assets?

We can think about the composition of Treasury holdings from the perspective of the Fed’s balance sheet alone or from the perspective of the consolidated government balance sheet.

Considering only the Fed’s balance sheet, efficiency considerations counsel against holding Treasury securities that are in particularly scarce supply. For example, the Fed has usually limited its holdings of any one security and avoided buying in-demand securities such as those that are newly issued or cheapest to deliver into futures contracts. If long-term Treasury securities have a special liquidity premium, efficiency could also suggest steering away from holding them—but, as I will discuss in a moment, the consolidated government perspective undoes this conclusion.

Effectiveness, meanwhile, suggests the Fed should structure asset holdings so they normally remain in the background and do not distract from our primary monetary policy tool, the overnight rate target. While fluctuations in net income do not affect the Fed’s ability to conduct monetary policy, they can become a communications challenge. Roughly matching the duration of our assets and liabilities would reduce these fluctuations and could, thus, enhance the effectiveness of policy communications. The Fed’s liabilities consist primarily of currency, which pays zero interest and has very long duration, and reserves, which pay a floating rate and have zero duration. Asset-liability matching thus suggests holding a mix of short-term and long-term securities. A portfolio that’s neutral relative to the outstanding Treasury universe would be one way to approximate this.

Some observers argue that tilting the portfolio toward shorter-duration assets would provide flexibility that could enhance effectiveness. Short-dated assets can run off to rapidly reduce reserve supply. Or the central bank can trade them for longer-dated assets to absorb duration and influence term premiums without expanding reserve supply. These forms of flexibility were valuable in the pre-GFC scarce-reserves implementation regime, in which we needed to maintain precise control of reserve supply. And, indeed, we tilted the Fed’s portfolio toward Treasury bills in that regime. But the ample-reserves regime reduces the need for precise control of reserve supply. Our balance sheet has also grown, so even a neutral portfolio would now contain a substantial dollar quantity of Treasury bills. In the current regime, then, I don’t think tilting the steady-state portfolio toward short-dated assets would materially enhance effectiveness.

Importantly, Treasury securities the Fed holds as assets are liabilities for the Treasury Department. They cancel out on the consolidated government balance sheet. As a result, the consolidated perspective can shed a different light on what’s efficient and effective. Let me give a simple example. Suppose securities of a certain duration have a particular liquidity value in the market. You might think it’s inefficient for the Fed to hold such securities, since the market values them highly. However, the Treasury could conceivably issue more such securities to meet the market’s demand. So, from a consolidated perspective, the Fed’s actions aren’t the sole determinant of efficiency.

In the U.S., the Fed sets monetary policy while Congress and the Treasury determine the fiscal costs taxpayers support. One of those fiscal costs comes from the duration risk associated with potential interest rate fluctuations on government debt. The Treasury determines the maturity structure of the consolidated government debt and the associated duration risk. When certain securities have a particular liquidity value, for example, Treasury decides whether to supply more of them, taking account of how this action would affect taxpayers’ exposure to duration risk.

If Treasury can offset what the Fed holds, does the consolidated perspective imply anything at all for the Fed’s asset choices? I think it does. In practice, Treasury keeps its issuance regular and predictable to foster an attractive market for investors. Given this approach, I believe the most appropriate strategy for the Fed is to likewise be predictable by purchasing a neutral mix of durations relative to whatever mix Treasury chooses to issue. Then there is nothing for Treasury to offset. The Fed-only perspective also suggested a neutral portfolio. My bottom line, then, is that from both angles, in the U.S., a neutral portfolio would be efficient and effective.

At present, the Fed’s portfolio is significantly overweight longer-term securities and underweight Treasury bills. Our reinvestments of maturing securities represent a small fraction of the portfolio. And when we eventually reach an efficient level of reserves and need to begin expanding the portfolio in line with growth in demand for our liabilities, those purchases will also be small relative to existing holdings. In this context, it could take many years to reach a neutral mix of holdings by structuring our purchases to be proportional to issuance. So although I view a neutral mix of purchases relative to issuance as appropriate in the long run, it would make sense in the medium term to overweight purchases of shorter-dated securities so as to more promptly return the Fed’s holdings to a neutral allocation.

Finally, while securities represent nearly all the Fed’s assets, we’ve always held small quantities of other assets, such as discount window loans and repos. Loans also feature prominently among the assets of other central banks.

In the long term, in my view, it could be interesting to consider whether allocating a modest share of the Fed’s long-run balance sheet to loans or repos could improve the efficiency and effectiveness of policy implementation. For example, auctioning a fixed quantity of discount window loans each day could encourage banks’ operational readiness and demonstrate that borrowing is a normal activity for healthy firms.[14]

Such a facility might also smooth the redistribution of reserves around the banking system. The U.S. has about 9,000 banks and credit unions. Unexpected payment shocks can leave some of them short of reserves at the same time as others have a surplus. Frictions in interbank lending may slow the movement of reserves to whichever banks need them most. But if the Fed held a daily lending auction, the depository institutions most in need of reserves on any given day would likely place the highest bids, automatically redistributing liquidity away from firms with less need.

Let me just strongly emphasize that the FOMC is not considering any changes to its implementation framework, and even beginning to consider such a tool would require substantial conversation, analysis and learning from the experience of other central banks.

Conclusion

To conclude, central banks naturally design their balance sheets and policy implementation frameworks to reflect institutional features of the jurisdictions in which they operate. The resulting frameworks vary on the surface, but there are commonalities in the underlying policy principles. I’ve argued today that central bank balance sheets should be efficient and effective. Those twin principles imply keeping money market rates close to the interest rate on reserves, supplying ample reserves and maintaining strong ceiling or lending tools. I find it heartening that central banks around the world have converged on these fundamental approaches, and I hope we will continue to learn from one another, as we have done the past two days.

Thank you.

Notes

I am grateful to Sam Schulhofer-Wohl for assistance in preparing these remarks.

1. Federal Open Market Committee, “Policy Normalization Principles and Plans,” Sept. 16, 2014.
2. Federal Open Market Committee, “Principles for Reducing the Size of the Federal Reserve’s Balance Sheet,” Jan. 26, 2022; Federal Open Market Committee, “Plans for Reducing the Size of the Federal Reserve’s Balance Sheet,” May 4, 2022.
3. Lorie K. Logan, “Ample reserves and the Friedman rule,” remarks before the European Central Bank Conference on Money Markets, Nov. 10, 2023.
4. The Friedman rule (Milton Friedman, “The Optimum Quantity of Money,” in The Optimum Quantity of Money and Other Essays, Chicago: Aldine, 1969) calls for equating the opportunity cost of holding money to the marginal social cost of creating money. When money does not bear interest, the Friedman rule also has implications for the optimal inflation rate, which are beyond the scope of these remarks. See Logan (2023).
5. See, for example, Mark E. Paddrik, Carlos A. Ramirez and Matthew J. McCormick, “The Dynamics of the U.S. Overnight Triparty Repo Market,” FEDS Notes, Aug. 2, 2021.
6. Annette Vissing-Jorgensen, “Balance Sheet Policy Above the ELB,” remarks at the European Central Bank Forum on Central Banking, June 26–28, 2023.
7. In the Federal Reserve’s March 2024 Senior Financial Officer Survey, the median domestic bank respondent said it would not lend federal funds overnight for less than a 25-basis-point spread over interest on reserves. See also Kyungmin Kim, Antoine Martin and Ed Nosal, “Can the U.S. Interbank Market Be Revived?” Journal of Money, Credit and Banking 52(7), 1645-1689, October 2020.
8. See, for example, Simon Potter, “Money Markets at a Crossroads: Policy Implementation at a Time of Structural Change,” remarks at the Master of Applied Economics’ Distinguished Speaker Series, University of California, Los Angeles, April 5, 2017.
9. Board of Governors of the Federal Reserve System, “Request for Information and Comment on Operational Aspects of Federal Reserve Bank Extensions of Discount Window and Intraday Credit,” Federal Register 89(175), 73415-73418.
10. Federal Reserve Bank of New York, “Statement Regarding Repurchase Agreement Small Value Exercise,” February 20, 2025.
11. See, for example, James A. Clouse, Sebastian Infante, and Zeynep Senyuz, “Market-Based Indicators on the Road to Ample Reserves,” FEDS Notes, Jan. 31, 2025.
12. Elsewhere, I have discussed which money market rates I am watching most closely to assess liquidity conditions. The secured overnight financing rate, while an important reference rate, includes some transactions whose rates reflect compensation for intermediating funds from the triparty segment of the repo market to cash borrowers who lack triparty access. For this reason, I believe the somewhat narrower tri-party general collateral rate currently provides a cleaner read on liquidity conditions in U.S. secured markets. See Lorie K. Logan, “Normalizing the FOMC’s monetary policy tools,” remarks at the Securities Industry and Financial Markets Association annual meeting, Oct. 21, 2024. 
13. Federal Open Market Committee, “Principles for Reducing the Size of the Federal Reserve’s Balance Sheet,” Jan. 26, 2022.
14. Such an offering would be similar to the Federal Reserve’s Term Auction Facility, which offered term discount window loans at an auction rate in 2007 through 2010.