Opening remarks for panel titled ‘Post-Pandemic Challenges for Monetary Policy Implementation’

It is an honor to participate in this conference commemorating the 100th anniversary of the founding of the Banco de México. Gov. [Victoria] Rodriguez Ceja, thank you very much for inviting me. Let me note at the outset that the views I share are mine and not necessarily those of my colleagues on the Federal Open Market Committee (FOMC).

The Banco de México is a leader in the global central banking community. In 1994, the bank gained policy independence, which strengthened its ability to contain inflation. For the people of Mexico, that outcome has supported economic and financial stability. And for the world, it’s provided a strong example of the value of independent, objective monetary policymaking.

The bank’s thought leadership receives worldwide recognition. Banxico is a driving force behind both CEMLA, the Center for Latin American Monetary Studies, and the Bank for International Settlements Representative Office for the Americas, both based here in Mexico City.

I greatly appreciate the warm relationship between Banxico and the Dallas Fed. The economic ties between Mexico and the Eleventh Federal Reserve District are deep and longstanding. Our two banks co-host an annual conference on international economics, along with the University of Houston. And I and our entire research and leadership teams learn a great deal from the ongoing dialogue between our institutions.

It is a true pleasure to celebrate your centennial with you here today and to have the opportunity to discuss the challenges of monetary policy implementation after the pandemic. This is an excellent topic for a centennial conference examining the most crucial questions in central banking. Implementing a central bank’s monetary policy efficiently and effectively is vital to the credibility of that policy and to public trust in the central bank. Yet there is no obvious single right way to implement policy. Every central bank’s operating framework reflects its history, its unique institutional context and the market environment it faces. And we learn over time. In light of both the extraordinary policy implementation actions many central banks undertook in response to the pandemic, and the changes in financial markets in recent years, now is a fitting time to take stock.

Balance sheet normalization

In my view, a key current challenge of monetary policy implementation is how to return central bank balance sheets to more normal sizes and compositions following the expansions undertaken in response to the pandemic. I’ll focus my discussion on G7 central banks such as the Fed, European Central Bank (ECB), Bank of England (BoE) and Bank of Canada (BoC), which have similar approaches to monetary policy implementation and operate in broadly similar financial systems. But the challenge of balance sheet normalization applies much more broadly, as central banks across a wide range of economies expanded their balance sheets earlier this decade. I hope our conversation today will tackle approaches to normalization around the world.

Normalization affects both sides of a central bank’s balance sheet. Let’s start with liabilities.

Most central banks’ main liabilities are currency and bank reserves. Currency demand moves slowly, so shrinking the balance sheet typically involves reducing the quantity of reserves. The trick is how to do so without creating undue upward movement in overnight interest rates, which are the price of reserves.

The Fed, ECB, BoE and BoC all pay interest on reserves, and all intend to operate in the long run in a framework in which money market rates are close to the interest rate on reserves. That arrangement implements what’s known as the Friedman rule, because it minimizes banks’ opportunity cost of holding reserves. As I’ve argued elsewhere, following the Friedman rule enhances the efficiency of the banking and payments systems and supports both financial stability and effective rate control. I will refer to this type of arrangement as a regime of ample reserves. I want to emphasize that I’m intentionally not calling it a floor system, even though market rates will typically be close to the floor established by the interest rate on reserves. The central banks I’ve mentioned all also maintain lending or funding facilities that help put a ceiling on rates, such as, in the United States, the discount window and the Standing Repo Facility (SRF). Ceiling tools can play an important role in policy implementation both during and after balance sheet normalization. What defines these systems is a supply of reserves ample enough to keep market rates near interest on reserves—not that rates are necessarily pushed to the floor.

To keep money market rates close to interest on reserves, the central bank must supply at least as many reserves as commercial banks demand with that configuration of rates. Otherwise, money market rates will rise above the central bank’s policy target.

That sounds simple, but it raises several questions. How should the central bank supply the reserves? How should the central bank know how many reserves to supply? And what do we even mean by the quantity of reserves that commercial banks demand?

The last question holds the key to the other two.

On any given day, any given commercial bank has some quantity of reserves it would like to hold as a function of the spread between market interest rates and interest on reserves. When market rates rise, the bank will want to hold fewer reserves and invest more in market instruments. But a bank has limited options for adjusting its reserve holdings on any given day. It must maintain enough reserves to manage whatever liquidity risks are inherent in its current business model. If it has more than enough reserves for its liquidity needs, it could sell or lend funds to another market participant, but some of those reserves might boomerang back in the form of deposits.

The bank has more options for reducing its reserve demand in the long run. It can cut the interest rates it pays on deposits, so some depositors may take their funds elsewhere. And it can reconfigure its business to require less liquidity.

Moreover, there can be distributional frictions in the banking system. When reserves drop sharply on a single day, the banks that initially lose reserves aren’t necessarily those that value reserves the least. Until the reserves redistribute, there’s a temporary increase in aggregate reserve demand.

Bottom line: as aggregate reserve levels come down, commercial banks’ short-run demand for reserves will likely be higher and less responsive to interest rate spreads than the long-run demand. When a central bank seeks to meet commercial banks’ demand for reserves, it must decide whether that means the short-run demand or the likely lower long-run demand.

In my view, it is preferable to aim in the long run to meet the long-run demand. Otherwise, there can be a ratchet effect.

Since the Global Financial Crisis, central banks have repeatedly bought assets to support financial stability or stimulate the economy when overnight rates were at the effective lower bound. Particularly when stimulus is the goal, those newly acquired assets need to stay on the central bank balance sheet for some time to generate the intended benefits. But the additional reserves that fund the assets also stick around for some time. Those additional reserves tend to push market rates meaningfully below the interest rate on reserves, creating an environment where reserves are more than ample—perhaps abundant or super-abundant. And commercial banks can get used to that environment. They find it profitable to expand their deposit bases or develop business models that rely on abundant liquidity. The short-run demand curve for reserves shifts outward. Unless the central bank finds a way to return to the old long-run demand for reserves, the elevated short-run demand will become the new long-run demand. And instead of the balance sheet expansion being a temporary response to economic or financial stress, it will become permanent. Repeat that over multiple stress episodes, and you’d have a recipe for ever-expanding central bank balance sheets and no route back to normal.

To get back to the long-run demand curve for reserves, central banks need to bring the average level of market rates up closer to interest on reserves in a sustained way. That will give commercial banks an incentive to make the long-term adjustments that can bring reserve demand down. Importantly, market rates can fluctuate from day to day. U.S. repo rates, for example, have recently varied in a range of 10 or so basis points. So, moving the average level up closer to interest on reserves also requires some tolerance for temporary, modest moves above interest on reserves.

With a sustained shift up in market rates as the goal, there are two basic ways a central bank can identify how far to reduce reserves: ceiling tools and glide paths.

If the central bank has confidence that market participants are willing and able to access its ceiling tools, it can simply reduce reserves until the point that ceiling tools see routine use. This is basically the approach that I interpret the ECB and BoE to be following. If the economic cost of funding from the ceiling tools modestly exceeds the interest rate on reserves, banks will have an incentive to bring reserve demand back down to the long-run level. The spread between interest on reserves and the cost of funds from ceiling tools can come either from a spread between official interest rates or from other costs such as collateral haircuts.

Let me emphasize two necessary conditions for this approach to maintain rate control. Market participants must be willing and able to access the ceiling tools when market rates exceed the ceiling tool rates. If ceiling tools are stigmatized or operationally difficult to access, market rates could spike too high. Additionally, the ceiling rates must not exceed interest on reserves by too much. Otherwise, market rates will have too much room to rise before firms will borrow from the ceiling tools.

A complementary strategy is for the central bank to come in for a gradual landing on reserves: a glide path. By laying out a smooth path to lower reserve levels, the central bank gives market participants time to adjust to an environment of lower reserves. Along that path, ceiling tools can still contribute to rate control. And, to reach the long-run level of reserve demand, the central bank must still have some tolerance for modest, temporary rate pressures that can occur as the average level of rates moves toward interest on reserves and creates an incentive for banks to shed reserves. But the evolution of money market rates will provide important signals about when reserves are approaching the right level. The Fed and Bank of Canada have both employed what I see as variants of this strategy, and it is working well.

The glide path does not rely as much as the first strategy on confidence in the efficacy of ceiling tools. At the Fed, while we have made strides in enhancing the effectiveness of the discount window and Standing Repo Facility, it would be worthwhile to consider further steps, such as increasing or removing limits on the SRF’s size or centrally clearing those transactions. So, I believe bringing reserves down gradually, while also making our ceiling tools available and encouraging market participants to use them when they are economically attractive, will be an effective strategy in the United States.

In the U.S., repo rates have averaged about 8 basis points below interest on reserves in recent months. That tells me we have more room to reduce reserves. We could see some temporary pressure around the tax date and quarter-end in September. I was encouraged to see market participants using the SRF over the June quarter-end, and I anticipate they will similarly use our ceiling tools if necessary in September. That will allow us to continue gradually bringing reserves to a more efficient level with market rates close to, but perhaps slightly below, interest on reserves on average over time.

Long-run reserve supply

Once the central bank completes the normalization process and reaches the long-run level of reserves, it must decide how to supply reserves in steady state. What fraction of reserves should be backed by outright asset holdings, and what fraction should come from ceiling tools? The Friedman rule doesn’t answer this question, because either approach can, in principle, keep market rates close to interest on reserves. Rather, one might turn to considerations of footprint and rate control.

It’s often asserted that a central bank should avoid having too large a footprint in financial markets. At a high level, that makes sense to me. A market economy should rely mainly on private-sector activity. But what does footprint mean? One view is that a central bank’s footprint relates to the volume of its transactions with the private sector or the number of counterparties it transacts with. In that case, the central bank can minimize its footprint by backing reserves with relatively stable outright holdings of government securities, while making relatively few loans to market participants through its ceiling tools and conducting relatively few operations to offset changes in autonomous factors that add or drain reserves. But another view is that a central bank’s footprint relates to the size of its balance sheet. If a central bank wants to minimize transactions and loans through ceiling tools, it will need to supply a buffer of extra reserves to account for uncertainty in estimates of reserve demand and absorb variability in autonomous factors. So, if you think footprint relates to balance sheet size, the central bank should supply more reserves through ceiling tools (potentially to a broader range of direct counterparties) and temporary operations and less from outright holdings.

How best to define footprint depends on a country’s economic, financial and institutional context. I don’t think there are universal answers that apply to all central banks. Forums like this one present valuable opportunities to learn from each other.

Rate control is a second consideration. Just as in the balance sheet normalization process, a central bank can support rate control in the long run both by offering ceiling tools and by smoothly providing enough outright reserves to reliably meet banks’ demand. If market participants are willing and able to access ceiling tools, and if the spread between interest on reserves and lending rates is not too wide, the ceiling tools will effectively cap money market rates. If the central bank prefers a wider rate spread, or if there are concerns about the usability of ceiling tools, it becomes more important to smooth reserve supply. That can require forecasting large swings in autonomous factors that add or drain reserves and preplanning purchase operations to offset them, as described in a 2019 staff memo to the FOMC. I don’t view such operations as fine-tuning reserve supply to control money market rates the way the Fed did before the Global Financial Crisis. They simply ensure a smooth and ample supply of reserves that provides a predictable operating environment for banks.

Even if the aggregate supply of reserves is stable, payment shocks can shift substantial amounts of reserves from one bank to another on any given day. Unless the banking system efficiently redistributes reserves among banks, those shocks can temporarily raise the aggregate demand for reserves, because some reserves will be trapped at banks that don’t particularly value them. A central bank could reduce the long-run demand for reserves and enhance rate control by providing ceiling tools that help redistribute reserves. For example, if the central bank held a daily auction of a fixed quantity of discount window loans, those reserves would naturally flow to the banks that value them most.

Assets

Let me conclude by briefly discussing the asset side of the balance sheet. I’ve already mentioned some trade-offs between backing reserves with outright asset holdings versus loans through ceiling tools. For whatever part of its balance sheet consists of outright assets, a central bank faces additional choices about which assets to hold. While the details of these choices will necessarily depend on the specifics of a central bank’s institutional structure and market environment, a couple of principles seem to me to be relevant more universally.

First, I believe a central bank’s outright asset holdings should primarily comprise whatever serves as the reference asset class in its economy. In that way, the central bank’s holdings will primarily influence broad, baseline financial conditions, rather than spreads between different markets. Treasury securities are the reference asset class in the United States. The Treasury yield curve is an economy-wide benchmark, and the market for Treasuries is as deep and liquid as they come. So it makes sense for the FOMC to hold primarily Treasuries in the long run, as we’ve repeatedly said we intend to do. (Currently, on a par basis, the Fed’s securities holdings consist of about 63 percent long-term Treasury securities, 3 percent Treasury bills and 34 percent agency mortgage-backed securities that are effectively guaranteed by the U.S. government.) In economies where the supply of government debt is smaller, or for a multinational central bank like the ECB, the reference asset could be different.

A second principle is that, in selecting its core asset holdings, a central bank should respect the boundary between fiscal policy and monetary policy. The nature of that boundary varies across institutions, and the relevant considerations will also depend on what type of assets the central bank primarily holds. For the Fed, the question here is which Treasury securities to hold—in particular, which durations.

The Fed remits its net income to the Treasury. From an income perspective, while our choice about which Treasury securities to hold as assets can change our remittances, it’s a wash for taxpayers because the interest income we receive on Treasury holdings represents an interest expense for the Treasury Department.

However, to the extent that Treasury doesn’t adjust its issuance to offset the Fed’s holdings, changes in the maturity composition of the Fed’s assets can change the maturity structure of the consolidated public debt. In turn, that maturity structure influences the amount of duration risk that taxpayers bear from potential interest rate fluctuations on government debt. In my view, and as I’ve said before, the maturity structure of the consolidated government debt in the U.S. is a choice for the Treasury Department, not the Fed. There are many conceivable ways to maintain that division of responsibilities. But a simple and easily communicated one would be for the Fed, in the long run, to simply make its asset purchases proportional to Treasury issuance. In the medium term, overweighting Treasury bills in our purchases could more expeditiously move our current mix of holdings closer to matching the market. The FOMC has made no decisions on the long-run composition of its Treasury holdings, and I look forward to continuing to discuss this topic with my colleagues.

Thank you.

Note

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