A typical central bank formulates policy in the very short run by choosing a target for some short-term (usually overnight) nominal interest rate. The procedure for hitting that target depends on the nature of financial markets, and the structure of the banking sytem, among other things. One approach is to operate a channel or corridor system, under which the central bank lends to financial institutions at x%, offers financial institutions deposits (reserves) at the central bank at y%, with a target overnight interest rate of z%. The channel is structured with x >= z >= y, and arbitrage prevents the overnight rate from escaping the channel. A good example of a channel system in operation is Canada, where x - z = z - y = 25 basis points. In the Canadian banking system there are no reserve requirements, and the financial system operates essentially with zero reserves (with a little slippage) overnight while the corridor system is in operation. The U.S. system before the financial crisis was a type of channel system with the interest rate on reserves at zero, the Fed's discount rate determining the upper bound on the overnight rate, and the fed funds rate serving as the target interest rate.
If there are sufficient reserves outstanding (and sufficient need not be much - see my Review article on what happened in Canada, 2010-2011), then a channel system becomes a floor system. That is, in the example, arbitrage should dictate that the overnight rate is y%. But, in the U.S. financial system, arbitrage is imperfect - there are frictions. First, government sponsored enterprises (GSEs) cannot receive interest on their reserve accounts with the Fed, by law. Second, the financial institutions that do receive interest on reserves - depository institutions - face regulatory costs to holding reserves, which is where the friction comes in. So, the Fed has another instrument - overnight reverse repurchase agreements (or ON-RRPs) - which helps it to hit its target for the overnight fed funds rate. An ON-RRP is an overnight loan to the Fed, so think of this as just another interest-bearing Fed liability, like reserves. There is an expanded list of counterparties who can engage in ON-RRPs with the Fed, and that list includes GSEs and money market mutual funds - the first set of institutions cannot receive interest on reserves, and the second do not have reserve accounts. So, the Fed's ON-RRP facility extends the reach of interest-bearing Fed liabilities.
With the ON-RRP facility in place, U.S. monetary policy implementation is unique in the world, as far as I know. What, under ideal conditions, would be a floor system (given the large quantity of reserves outstanding) is actually a floor with a sub-floor. The Fed's discount rate (or "primary credit rate") is currently set at 1.0% which, if there were zero excess reserve outstanding, would normally determine an upper bound (roughly) on the fed funds rate. The floor is the interest rate on excess reserves, or IOER, which is currently set at 0.5%. The sub-floor is the ON-RRP rate, currently set by the Fed at 0.25%. The December 16 FOMC statement states that:
...the Committee decided to raise the target range for the federal funds rate to 1/4 to 1/2 percent.Information on the details of the directive that the Open Market Desk at the New York Fed received is in this implementation note. Basically, the idea is that the only operations that are necessary for the New York Fed for the immediate future are daily ON-RRP operations. These will be conducted at a rate of 0.25%, with no constraints other than available collateral - the securities on the Fed's balance sheet that are not otherwise accounted for - and per-counterparty limits of $30 billion.
Actual ON-RRP transactions are reported on the New York Fed's website. After the change in policy, the takeup on the ON-RRP facility on Thursday, Friday, and Monday, was, respectively, $105 billion, $143 billion, and $160 billion. So, the quantity of outstanding ON-RRPs has been increasing, but is very modest relative to reserves (at $2.5 trillion) or the quantity of available collateral, which is about $2 trillion. But did the New York Fed actually succeed in controlling the fed funds rate within the specified range of 0.25-0.50%? You can see the results here. On the first three days after liftoff, the effective fed funds rate (an average of rates in individual trades on the market) was 0.37%, 0.37%, and 0.36%, respectively, which is, perhaps surprisingly, in the middle of the range. The dispersion in fed funds rates is small (the standard deviation is 0.05), but there are trades on the fed funds market as high as 0.59% (above the IOER), and as low as 0.25% (the ON-RRP rate).
There can be dispersion in fed funds rates because of idiosyncratic counterparty risk in the market (fed funds is unsecured credit) - this was an important factor during the financial crisis for example. But, under current conditions, any dispersion in rates is in part due to the heterogeneous nature of trading. In particular, there are basically two kinds of borrowers in the market. The first is a depository institution which receives interest on reserves, and is borrowing from some financial institution - typically a GSE - that does not want to be caught holding reserves overnight that earn zero interest. These trades typically occur at an interest rate less than the IOER. The second is a depository institution which, much as in pre-financial crisis times, finds itself short of reserves at the end of the day, and borrows on the fed funds market to replenish reserves. Due to the huge stock of reserves currently in the system, there are very few of these depository institutions at the end of each day, and therefore very few of these transactions. Seemingly, these are the trades that occur above IOER, because the lenders face different opportunity costs than in cases where the lenders cannot receive interest on reserves.
The second stage of normalization will be continued increases in the range for the fed funds rate, followed by a reduction in the size of the Fed's balance sheet. Currently, the size of the Fed's asset portfolio is held constant, in nominal terms, through reinvestment in mortgage backed securities and Treasury securities as these assets mature. The Policy Normalization Principles and Plans state that
The Committee expects to cease or commence phasing out reinvestments after it begins increasing the target range for the federal funds rate; the timing will depend on how economic and financial conditions and the economic outlook evolve.As well,
The Committee currently does not anticipate selling agency mortgage-backed securities as part of the normalization process, although limited sales might be warranted in the longer run to reduce or eliminate residual holdings. The timing and pace of any sales would be communicated to the public in advance.So, balance sheet reduction, when it happens, will result when the the Fed's assets are maturing, and they are not being replaced. No outright sales, other than for fine-tuning purposes, are anticipated. A normal balance sheet will have: (i) a small quantity of reserves, on the order of what was outstanding prior to the financial crisis; (ii) a portfolio consisting of only Treasury securities (no mortgage backed securities, for example); (iii) an asset portfolio with a shorter average maturity than currently, again comparable to what existed before the financial crisis. According to this paper by Carpenter et al., normalization in the size of the balance sheet might take 6 years or more from the time reinvestment ends, and normalization in terms of average maturity of the asset portfolio will take even longer.
As normalization proceeds, an issue will arise as to how balance sheet reduction relates to increases in the fed funds rate range. Clearly, the large-scale asset purchases that occurred were viewed by some policymakers as being equivalent to reductions in the fed funds rate target. One view is that we can translate a quantity of nominal asset purchases by the Fed into a given basis point decrease in the fed funds rate, were such a decrease feasible. A second view is that the size of the balance sheet is immaterial - what matters is the composition of the Fed's asset holdings, for example as measured by average maturity. A third view is that neither the size of the balance sheet nor its composition makes any difference - quantitative easing is neutral, and only the short-term nominal interest rate matters. Which view one takes clearly matters a great deal for how normalization should proceed.
I wanted to think about some of these issues, and understand more about how the Fed's ON-RRP facility works, so I wrote this paper. This is a model with two banking sectors - basically regulated and unregulated. In the regulated sector, banks can hold interest-bearing reserves, and they have a capital requirement. Unregulated banks cannot hold reserves, but they also don't have capital requirements. There are four assets: currency, reserves, government debt (one period nominal bonds), and private assets. Regulated and unregulated banks serve different clienteles. The regulated banks offer deposit contracts that provide for currency withdrawal on demand and transactions services (looks like the opportunity to do debit card transactions), and the unregulated banks provide intermediation services that can be interpreted as involving repurchase agreements (repos) using government debt as collateral. In the model, currency and government debt have special roles: currency is the only asset accepted in some transactions, and government debt is the only collateral acceptable in some types of credit transactions. There is an interbank credit market on which regulated and unregulated banks can trade.
In the model, if the balance sheet of the central bank gets large enough, then there can be a positive margin between the interest rate on reserves and the interbank interest rate. As well, there can be a positive margin between the interbank rate and the interest rate on government debt, provided government debt is in sufficiently low supply, in a well-defined sense. This corresponds to what we have been observing recently. In particular, in the last few days short-term T-bills have been trading not only below the interest rate on reserves and the fed funds rate, but below the ON-RRP rate. Basically, the interest rate on government debt in the model can be lower than the interbank rate due to a higher liquidity premium on government debt.
I introduce an ON-RRP facility in the model (basically the unregulated banks can hold interest-bearing central bank liabilities). This does what it is supposed to do - it puts a floor under the interbank rate. As well, more ON-RRPs are always welfare improving. Further, a large central bank balance sheet is a bad thing. Reserves in the model are assets that sit on the balance sheets of regulated banks, and are costly to hold because of the capital requirement. That is, because of the capital requirement, a swap of reserves for government debt tightens collateral constraints and can make everyone worse off. As well, the expanded balance sheet converts assets that are useful in particular transactions as collateral (government debt) into reserves, which are not useful in that sense.
In terms of "tightening" by way of an increase in the interest rate on reserves vs. a reduction in the balance sheet, these two policy changes have similar effects on market interest rates. However, the effects on quantities and on welfare are very different.
I think these results are interesting. Let me know what you think.