We describe a framework for understanding the size and composition of the normalized balance sheet.1
The Committee sees a number of advantages to a policy implementation framework, like the current one, in which reserves are abundant and the Fed doesn’t need to actively manage the supply of reserves to achieve the target policy rate, as had been the practice until late 2008.
▪ Monetary policy is easier to implement in this regime: Interest paid on reserves, as well as the reverse repo facility, form a floor for the funds rate, while the abundance of reserves ensures that the funds rate won’t trade much higher than that floor.
▪ In addition, participants have pointed out that a regime with abundant reserves would allow the FOMC more flexibility in implementing policy in different circumstances, for example, when the funds rate is at its effective lower bound.
On the other hand, FOMC participants also think that the balance sheet can still deliver those benefits even if it is “much smaller” than it is today. And some participants see potential costs of maintaining a balance sheet that is larger than operationally necessary.
▪ The larger the balance sheet, the greater the Fed’s holdings of Treasuries (and perhaps MBS) and the greater the risk that the Fed would have to temporarily suspend remittances to Treasury as interest rates rise.
▪ In addition, the FOMC may see a smaller balance sheet as increasing its scope to increase holdings later through asset purchases, if necessary.
Participants also recognized that to operate in such a regime the balance sheet would likely have to be at least “somewhat larger” than the pre-crisis balance sheet because of the “trend growth of balance sheet items such as currency as well as a larger supply of reserves.”
▪ The key step in determining the size of a normalized balance sheet is adjusting for the growth of currency, a larger supply of reserves, and other balance sheet items.
▪ These adjustments leave the normalized balance sheet quite large, much larger than the pre-crisis balance sheet.
In coming months, the FOMC will announce the first steps in its plan for the normalization of the Fed’s balance sheet. That will entail the announcement of the timing of a change in reinvestment policy and the start of the runoff of Treasuries and MBS from the portfolio. We expect the current reinvestment policy (reinvestment of the full amount of principal) to be changed this year, though reinvestments will likely be phased out gradually over a period of perhaps six months to a year. The FOMC may also provide more explicit guidance about the longer-run size and composition of the normalized balance sheet, though currently it is unclear whether that guidance will be issued at the same time as the announcement about the change in reinvestment policy or sometime later.
The minutes of the November 2016 FOMC meeting provided some guidance about the size of the normalized balance sheet, but did not provide any numbers. There appeared to be a clear consensus among FOMC participants that it was better not to return to a system that involved active management of the supply of reserves. Rather, they preferred to continue to implement policy in a regime like the current one, which could be done with a balance sheet that was “much smaller” than at present, though “likely at least somewhat larger” than in the years before the financial crisis, reflecting “trend growth of balance sheet items such as currency as well as a larger supply of reserves.”
The Current Operating Framework
In the pre-crisis period, the FOMC implemented policy with a very small amount of reserves in the banking system. For example, in March 2007, there were just $16 billion in reserves. In this framework, the New York Fed needed to make only small adjustments to the level of reserves through open market operations to accurately hit the fed funds rate target.
During the financial crisis, the Fed’s emergency facilities drove up excess reserves, and the New York Fed no longer was able to manage the funds rate merely by adjusting the level of reserves. So the Fed began paying interest on reserves, creating a (soft) floor for the funds rate. After that, the FOMC implemented a number of asset purchase programs that sharply further increased the level of reserves and the size of the balance sheet. To strengthen the floor for the funds rate and the transmission from the funds rate to other short-term market rates, and to ensure that it would be able to manage the funds rate smoothly when it eventually wanted to raise the funds rate target off the effective lower bound, the Fed added the overnight reverse repo (ON RRP) facility, available to a wider set of counterparties.2 The Fed refers to the current operating framework as a “floor” system, where the interest rate on excess reserves (IOER rate) and ON RRP rates together set a floor, despite the outsized level of reserves, allowing the FOMC to stabilize the funds rate within a 25-basis-point range.3
2 An important reason why the funds rate had been trading so soft to the IOER rate is that, while GSEs traded in the fed funds market, they could not earn interest on reserves. GSEs now have access to the ON RRP facility and that reinforces the floor set by the IOER rate. The fact that a wide range of financial institutions have access to the ON RRP facility helps to strengthen both the floor and the transmission mechanism from the funds rate to other market rates. 3 The “floor” terminology may seem confusing, as the funds rate trades in a range set by the IOER rate and the ON RRP rate. This makes it look on the surface like a “corridor” system, the way most other central banks conduct monetary policy. In such a system, the corridor is set by the borrowing rate (the floor) and the lending rate (the ceiling). Reserves are then adjusted to keep the rate at its target within that corridor. The Fed does not have an active lending facility of the sort needed to create a hard ceiling in a corridor system. The discount window doesn’t qualify, in large part because of the stigma attached to borrowing there. While the IOER rate is higher than the federal funds rate, it is not a ceiling, in that it is putting upward pressure on the federal funds rate, not capping it. The ON RRP facility likewise puts upward pressure on the funds rate, reinforcing the soft floor created by IOER. While some FOMC participants may favor a corridor system, the floor system has so many advantages that we are quite confident they will retain this system. A corridor system still requires active intervention with open market operations to keep the policy rate near its target. A floor system allows a larger level of reserves and hence of safe assets. It also allows the FOMC’s balance sheet policies to be largely independent of interest rate policy, giving the FOMC the freedom to engage in quantitative easing.
The Advantages of a Larger Balance Sheet
The current framework is working well. Indeed, the Fed Board staff has told the Committee that there are important advantages of operating with a larger balance sheet than in the pre-crisis regime.
1. Ease of implementation: Maintaining a higher level of reserves while creating a floor for the funds rate by paying interest on reserves and offering reverse repurchase agreements (RRP) through a facility allows the FOMC to stabilize the funds rate in a narrow range without frequent interventions by the New York Fed and without reserve requirements.
2. Flexibility: The current regime provides an ability to implement policy under different economic circumstances. A number of participants said they viewed the expansion of the balance sheet by asset purchases as an important tool in situations where short-term interest rates were at the effective lower bound.
3. Correlation with market interest rates: The wider range of financial market counterparties participating in the RRP facility strengthens the relationship between the funds rate and other interest rates.
4. Liquidity: The current regime is more effective in providing liquidity in times of stress. The discount window, available only to depository institutions, proved inadequate given the stigma associated with borrowing from this source.
5. Market regulatory and structural changes: There is now greater demand for safe assets, including reserves, because of market and regulatory developments.
So perhaps one way of thinking about the size of the normalized balance sheet is how much it can be reduced and still provide these benefits.
The Costs of Too Large a Balance Sheet
There may also be costs of a balance sheet that is too large. First, there may be a constraint on the size of the balance sheet if the Fed wants to lower the risk of having to suspend remittances to the Treasury temporarily at some point as rates rise. This could happen if interest expenses exceed interest income at some point as the Fed raises short-term rates because interest paid on reserves and through the RRP facility increases more quickly than interest income from its asset holdings. A suspension of remittances would be politically awkward, to say the least. Second, starting with a smaller balance sheet would make the Committee more comfortable resuming asset purchases and expanding the balance sheet again in some future downturn.
The FOMC’s decision on the size of the normalized balance sheet will attempt to balance the benefits against the costs of a higher balance sheet.
The minutes indicated that the normalized balance sheet will likely be at least “somewhat larger” than the pre-crisis balance sheet because of “trend growth of balance sheet items such as currency as well as a larger supply of reserves.” Actually, we think it will be much larger than the pre-crisis balance sheet because these adjustments are very large.
For the hypothetical normalized balance sheet today, we use the actual level of currency, and then assume it grows at the rate of nominal income through 2023. That’s very straightforward, though one could quibble about the growth rate, given that currency grew at a much faster rate than nominal income from 2007 to 2017.
The adjustment that is the most important and hardest to judge is that of the target level of reserves, the amount of reserves necessary to keep banks’ demand for excess reserves satiated and keep the funds rate from rising significantly above its floor.
We start with the historical level of reserves in December 2008, the time at which the FOMC was transitioning from the pre-crisis to a new operating regime where it managed the funds rate by trying to set a floor with IOER. We interpret the level of reserves in December 2008 as the minimum level to allow the FOMC to implement policy at that time in the new regime. However, we adjust that down because we view the actual level in December 2008 as larger than normally would be necessary to implement policy in the new operating regime because of the elevated demand for liquidity at a time of very stressed markets. We call that level of reserves the “target” level, the level just necessary to implement policy in the new regime. We then assumed that the target level of reserves grew from this base at the rate of growth of nominal income. The adjustment for the presence of reverse repos is another very important one and this adjustment may also be subject to differences of views. As we noted, implementing monetary policy in the current operating regime requires the level of excess reserves to be high enough that the level of bank reserves no longer affects the funds rate. The level of reverse repos, on the other hand, depends on the demand for them by a wider set of counterparties at the prevailing reverse repo rate set by the Fed. The overnight reverse repo facility reinforces the floor. But the presence of reverse repos on the balance sheet does not change the level of reserves required to implement policy under the current operating regime. So, the adjustment for reverse repos is just to add them to the pre-crisis balance sheet adjusted for trend growth in currency and the supply of reserves.
Finally, we make assumptions about other items on the liability side of the balance sheet, such as the Treasury’s General Account.
What if the FOMC Wants a Smaller Balance Sheet?
These adjustments add up to something very large. They could easily leave the normalized hypothetical balance sheet in 2017 at $3 trillion or more. So, it is worthwhile to ask whether the FOMC might want to hold down the size of the normalized balance sheet, and if it did, how it could do so.
The case for taking steps to hold down the size of the normalized balance sheet comes back to the perception of costs associated with a balance sheet larger than necessary for operational efficiency, discussed above. Some of the costs might be perceived political risks, for example, political backlash to what Congress perceives as too large a balance sheet, on top of threats to the Fed’s independence that are already present for other reasons. Another reason could be the FOMC’s own discomfort with having such a large “normal” balance sheet, given the prospect that it might have to resort to asset purchases again in response to an adverse shock that forces the funds rate to the effective lower bound.
Where does the FOMC have the greatest ability to manage the size of the balance sheet? Not, of course, with the level and growth of currency. Somewhat with respect to the level of reserves and reverse repos, but likely at the cost of less effective control of the funds rate. Still, one way it could restrain the size of its normalized balance sheet is by managing the level of reverse repos. For example, it could limit the usage of overnight reverse repos by setting (binding) caps. The Fed might accept some cost in terms of operational efficiency to limit the size of the balance sheet in this way. It might also constrain the usage of reverse repos by foreign official and international accounts—usage of this long-standing facility has grown in recent years.
With the objective of ending up with a smaller normalized balance sheet, we assume a lower level for overnight reverse repos on the hypothetical normalized balance sheet in 2017 than on the balance sheet
today, and then assume they grow at the rate of nominal income growth. We also assume a lower level of reverse repos for foreign official and international accounts in the hypothetical balance sheet in 2017 compared to its level today, and then assume these also grow at the rate of growth of nominal income.
In the End, Experience Will Inform the Size of the Normalized Balance Sheet
In the end, the size of the normalized balance sheet will have a lot to do with the FOMC’s experience implementing monetary policy as it ends reinvestment and shrinks the balance sheet. Our next commentary will provide numerical estimates for the two normalized balance sheets (today and in 2023, when we assume the balance sheet will actually be normalized) using the framework developed here.
Our estimate for the hypothetical normalized balance sheet in 2017 is a little more than $2.5 trillion, and that grows to about $3.5 trillion in 2023.
Appendix. Supply and Demand and the Target Level of Reserves
In this Appendix, we use supply and demand curves for reserves to provide a qualitative illustration of the pre-crisis regime, the transition to the new regime, where we are today, and where the normalized balance sheet will be.
Figure 1 shows supply and demand curves for excess reserves to illustrate where the level of reserves was in late 2008 when the FOMC moved toward its current operating regime. This is a starting point for projecting the hypothetical target level of reserves today and in the future.
The supply curves for reserves are vertical because the Fed sets those levels. The demand curve for excess reserves by depository institutions depends on the interest rate and is downward sloping. It gets flatter and flatter until, beyond some level of demand, it becomes essentially flat. This is the point at which banks are said to become “satiated” with reserves.
Before the crisis, the Fed operated with a small quantity of reserves, so the supply curve intersected the demand curve on its steeply downward sloping part—S1 in Figure 1. In these circumstances, the New York Fed could hit the target level of the funds rate with small adjustments in the level of reserves. When the funds rate was lowered to a range of 0 to 25 basis points at the December 2008 FOMC meeting, the FOMC had reached the flat portion of the demand curve and the FOMC was in its new operating regime, which is still in place today. The level of reserves at that point is marked by the vertical line, S2, in Figure 1.
We therefore interpret S2 as the minimum level of reserves necessary to be on the flat portion of the demand curve. Today, following a series of asset purchase programs, the Fed is operating way out on that flat portion, shown by S3 in Figure 1. But that level of reserves is much larger than needed to be on the flat portion of the demand curve, so the question the FOMC is considering is where S2 is and just how far back toward S2 it should reduce the level of reserves.