Some have suggested that the government should increase the effective supply of public short-term debt in order to meet the demand by institutional cash managers for safe short-term debt. In the extreme case, if the government eliminated private issuance of short-term debt by issuing enough public short-term debt to satisfy this demand, then there could be no runs and no contagion. The same result could be achieved if the remaining short-term debt of the financial system was so small that runs would not be significant. These policies can be characterized as “crowding out” private short-term debt.
As suggested by Greenwood, Hanson, and Stein (2015), and Carson et al. (2014), the Treasury could crowd out runnable private short-term debt by replacing a certain amount of long-term Treasury debt issuance with shorter term Treasury debt. Investors that would otherwise buy and hold private short-term debt would instead buy Treasuries. This approach, of course, raises the questions of whether the Treasury has the capacity to do so and what the consequences would be if it did.
At the end of 2014, the total US debt outstanding was over $13 trillion (excluding intra-governmental holdings),1 $1.4 trillion of which was issued with a maturity of one year or less (“Treasury bills” or “T-bills”).2 T-bills are issued in various lengths: 4, 13, 26, or 52 weeks, and “cash management bills” can have maturities as short as a few days.3 The weighted average maturity of T-bills was 90 days between 1995 and 2009. Evidence suggests that T-bills with shorter maturity provide the most significant short-term safety and thus substitutability with short-term private debt.4 Thus, for purposes of crowding out private short-term debt, the US Treasury would have to shift its issuance to shorter maturities within T-bills.5 A potential issue to consider is whether such a shift would have significant consequences for the dollar’s status as a reserve currency. In February 2015, foreign holdings of US treasuries stood at roughly $4.1 trillion.6 Of this, only around $350 billion (8.5 percent of total) was held in short-term Treasury bills.7 If foreign governments have a preference for long-term debt, as appears to be the case, the shift to short-term issuance and corresponding decline in the supply of long-term Treasuries might weaken the dollar’s status as a dominant reserve currency. However, recent history has also illustrated periods of higher relative preferences for short-term US treasuries with many periods substantially higher than the 8.5 percent listed above. The percentage has peaked at nearly 28 percent since 2000,8 suggesting that a lowering of maturities may not significantly impact the dollar’s status as a reserve currency.
In figure 20.1 of chapter 20 we estimated that there was roughly $4.3 trillion in net runnable short-term debt in the United States at the end of 2014. The net amount is more relevant for this analysis than the gross amount, since the crowding out of the net amount will also eliminate the financial intermediation of short-term liabilities that is included in the gross amount, as discussed in chapter 20. Therefore it is only necessary for the government to crowd out the net amount of short-term liabilities. Assuming that $1 of T-bills crowds out $1 of private short-term debt, then the Treasury would have to issue $4.3 trillion in short-term liabilities—in addition to the $1.4 trillion outstanding—to completely crowd out private issuance. As a result $5.7 trillion (or 44 percent) of the Treasury’s $13 trillion total debt would have to be in T-bills in order to fully crowd out the $4.3 trillion private short-term debt market. Such a dramatic shift in the maturity structure of the Treasury’s liabilities would have to be an incremental process. The Treasury could not simply issue $4.3 trillion in short-term Treasury debt in the immediate term. This would, of course, increase the total outstanding debt and violate debt-ceiling limits that only moderately exceed total current debt levels. However, if the US Treasury were to fund all future projected fiscal deficits and pay off all outstanding debt as it matures with T-bills, then we estimate that it would reach $5.7 trillion in outstanding T-bills in 2017.9 Thus it is technically possible for this policy to fully crowd out private short-term debt issuance within a reasonably short time.
It is notable that even if the policy objective was a partial crowding out of private short-term debt issuance, then such a policy would still require a very large amount of increased T-bill issuance to meaningfully reduce contagion risk. As a counterfactual, if T-bill issuance were increased by $1 trillion, then this would only reduce the size of the private short-term debt market from $4.3 trillion to $3.3 trillion. A financial system with $3.3 trillion in private short-term debt is still quite vulnerable to contagion.
However, there are serious concerns with dramatically changing the maturity structure of US government debt that renders this approach impracticable. On the one hand, given a typically upward-sloping yield curve, lowering the maturity of debt issuance could reduce funding costs. On the other hand, switching government debt issuance to short-term maturities exposes the US government to the unpredictable fluctuations of future interest rates and forces the rollover of substantial amounts of debt at one time.10 This risk is highly problematic because it makes future budget planning, and thus future taxes, more unpredictable.11 Additionally it may be important for the United States to lock in low long-term interest rates today, during a period of historically low interest rates. For example, while 30-year rates today are approximately 2.7 percent, historically, they have been much higher, averaging as high as 9 percent in 1990 and 6 percent in 2000.12
In a political climate of debt ceilings and government shutdowns, the risk of increased funding costs is even more problematic. Although the Treasury has very little risk of becoming insolvent, there is a small but nonnegligible probability that a failure to increase the debt ceiling could force a temporary “technical default,” whereby the Treasury could not pay interest on all of its T-bills. During this time, it would be likely that the short-term creditors of the US Treasury would demand higher interest rates in order to bear this risk.13
The Federal Reserve could also “crowd out” private short-term debt, using its new tools of monetary policy—interest on excess reserves (“IOER”) and reverse repurchase agreements (“RRPs”). It is important to understand how these tools are used to conduct monetary policy in order to assess whether efforts to crowd out private short-term funding with these tools could conflict with monetary policy. This section first presents a very general overview of how and why the Fed uses IOER and RRPs for monetary policy. It then describes how these tools could be used to crowd out private issuance of short-term debt, and concludes with an evaluation of this approach.
The federal funds rate (“FFR”) is the rate at which banks lend reserves to one another overnight.15 This rate is determined by the market, although the Fed uses its tools of monetary policy to guide market rates to the target FFR announced by the Federal Open Markets Committee.16 The Fed establishes a baseline demand for reserves by setting reserve requirements, which are minimum amounts of reserves that depository institutions must hold against their deposits.17 Historically the Fed would set a target FFR and guide market rates to the target by marginally adjusting the level of reserves in the banking system.18 However, the total level of reserves and the overall size of the Fed’s balance sheet grew dramatically after the financial crisis. Presently banks hold approximately $90 billion in required reserves and approximately $2.6 trillion in excess reserve balances with the Fed.19 With the supply of reserves far exceeding banks’ required reserves, modest adjustments to the total outstanding level of required reserves would clearly no longer result in a meaningful change in banks’ ability or willingness to lend overnight at the FFR.20 This is because banks are easily able to meet moderately higher minimum reserve requirements in an environment where $2.6 trillion in excess reserves exist.21 As a result, while the Fed has historically used open market operations to set interest rates, recently rates are set by Fed guidance through announcements instead.22
It is important to note that the Fed created these excess reserves as part of their lending and buying programs during the financial crisis and quantitative easing programs after the crisis.23 Indeed only the Fed can increase or decrease the total level of reserves in the banking system, as the total level of reserves “is not affected by banks’ lending decisions.”24 While one bank may be able to marginally increase or decrease its level of reserves, this will necessarily be offset by an equal and offsetting increase or decrease in the reserves of another bank.25
Under current conditions the Fed must now use interest on excess reserves (“IOER”) and its “trial” reverse repo program (“RRP”) to set the effective floor of the FFR. IOER is simply an interest payment by the Fed on excess reserves. The Fed also pays interest on required reserves. An RRP is a short-term cash loan from an approved market participant (including nonbanks as well as banks), which is secured by collateral from the Federal Reserve’s securities portfolio.26 The Fed has implemented term RRPs of between one week and a month and overnight RRPs as part of their trial program. Current approved RRP counterparties include primary dealers, banks, MMFs, and GSEs.27 The Fed has stated that the program is temporary and for monetary policy purposes.28
These programs guide market rates toward the target FFR by eliminating the incentive for banks and nonbanks to lend at rates lower than what the Fed is paying on IOER and RRPs, respectively. This is because the interest paid on reserves and RRPs are generally viewed as risk free; why loan to anyone else at a lower rate when you can get at least the same rate from the Fed risk free?29 The IOER is currently set at 25 basis points.30 However, the IOER cannot set the effective floor for the FFR by itself, since nonbank short-term lenders like MMFs and the GSEs cannot hold reserves and are therefore willing to lend at rates below the IOER. The Fed’s RRP program sets the risk-free rate for these nonbank lenders. It is currently paying interest between 5 and 11 basis points.31 While banks have an incentive to arbitrage away any difference between the IOER rate and the RRP rate (having access to both), they cannot entirely do so, given the deposit insurance premiums paid to the FDIC on deposits. The Fed is widely expected to raise the FFR this year and has stated that it will use both IOER and RRP to do so.
The impact that IOER and the RRP program, respectively, have on the private short-term debt market differs in important respects. With regard to IOER, simply increasing the interest paid on reserves does not directly “crowd out” private short-term debt issuance because it only affects banks and the banking system as a whole cannot meaningfully increase the total level of reserves held at the Fed. However, academics have found that the RRP program clearly has a “crowding out” effect on private short-term debt issuance.32 Understanding why necessitates an explanation of how the RRP program affects the Fed’s balance sheet. According to a recently released NY Fed study on the RRP program, “RRP take-up does not expand the size of the Federal Reserve’s balance sheet … [it] shifts the composition of the Federal Reserve’s liabilities from [excess] reserves held by banks to RRPs that can be held by a wider range of institutions.”33 In other words, the amount of excess reserves decreases, as RRPs take up increases, and therefore the amount of total public (i.e., Fed) short-term liabilities stays neutral.
But if the total amount of public short-term liabilities does not increase, then why do RRPs “crowd out” private short-term debt issuance? According to the NY Fed, “the crowding out of private financing that results from greater use of RRPs largely represents a reduction in private lending by money market investors (with access to RRPs) to banks that are financing reserves.”34 Instead, these investors lend to the Fed.
This is best demonstrated with an example. For example, absent the RRP program, a money market fund would have been incentivized to enter into a repo with a bank or to buy commercial paper issued by a bank at a rate exceeding 0 percent. If the money market fund had previously deposited cash with the bank, then the switch to supplying funds to the Fed in the RRP would shrink the bank’s deposits. The private short-term liability issued by the bank (the repo, commercial paper, or deposit) would have clearly counted toward the overall $8.1 trillion in private short-term debt, and such private short-term debt overall has been reduced.
The size of the RRP program (and therefore the extent to which it crowds out private short-term debt) depends on the interest rates offered by the Fed.35 The Fed has found that whenever it offers rates that are slightly closer to prevailing short-term market repo rates—the RRP rate is generally lower than market repo rates—then it gets increased take up on its RRP program.36
This link between size of the RRP program and interest rates can also be clarified with an example. The current program, which is paying between 5 and 11 basis points, is not meeting its size caps. As of December 31, 2014, the Fed offered $600 billion in RRPs: $300 billion in overnight RRPs (“ON RRPs”) and $300 billion in term RRPs and had total RRP take up of only $397 billion—$171 billion in ON RRPs and $226 billion in term RRPs.37 Clearly, the Fed could increase uptake to the full $600 billion in RRPs by offering substantially higher rates (e.g., 1 percent). In this example money market investors that are lending at rates below 1 percent would be incentivized to enter into an RRP transaction with the Fed instead of lending to riskier private counterparties at a lower rate. However, raising the rate to 1 percent would of course be inconsistent with Fed’s monetary policy goals, which currently require near zero interest rates. In a very low interest rate environment, it is difficult to use RRP to crowd out.
But the size of the RRP program is expected to grow if and when the Fed raises rates. According to the NY Fed, when the FOMC decides to raise the target FFR, the extent to which the RRP program “crowds out” private short-term debt will also increase.38 Indeed, if the RRP program is to set the effective floor on the FFR, then the growth of the RRP program is likely. This is because in order to set the floor of the FFR, the RRP program must offer a sufficient supply of RRPs to meet the demand for RRPs at that interest rate.39 Otherwise, money market investors will still be incentivized to lend in private markets at a rate below the interest rate paid by the RRP program. However, the extent to which the size of the RRP program will grow as the Fed raises rates is less clear (especially since the Fed has stated it is a temporary program). Certain market participants have estimated that the RRP program could grow to $1 trillion.40 At the same time, while it is likely that the Fed must grow the RRP program to increase rates, there is a possibility that the Fed can announce rates rather than expand its balance sheet to set interest rates.41 The Fed can also use both the RRP and the IOER to set interest rates wherever it wants.
Banks currently hold $2.6 trillion in excess reserves with the Fed. Therefore the size of the RRP program could be increased to $2.6 trillion without the Fed having to create reserves by buying additional assets or lending. However, the relationship between the quantity of uptake on RRPs and the interest rate that must be offered on RRPs to achieve that uptake will determine whether RRPs can be used to crowd out private short-term debt issuance without conflicting with monetary policy goals.
Once again, an example is instructive. Suppose that the Fed wanted to set a floor interest rate of 1 percent in order to achieve its monetary policy objectives. Suppose further that the Fed, completely apart from its monetary policy objectives and in an effort to foster financial stability, also wanted to crowd out $500 billion of private issuance. If the take-up at 1 percent were less than $500 billion, the Fed would have to raise the RRP rate to achieve the desired level of crowding out. If only very minor adjustments to the rate paid on RRPs (0.01 percent) were necessary to achieve large increases in uptake (hundreds of billions), then the Fed would be able to substantially increase uptake without conflicting with monetary policy goals. However, if larger adjustments were necessary (0.25 percent or greater) to obtain a meaningful increase in uptake, then achieving an increase in size of the RRP program would clearly conflict with monetary policy goals. Note, however, that this may be less of a concern in practice because the financial markets will observe the RRP rates and adjust market rates accordingly.
The extent of uptake would depend in part on the spread between the Fed RRP rate and the rates paid in private repo markets. For example, if money market investors are collectively lending trillions at a rate only 0.01 percent above the Fed RRP rate, then it is likely that a very marginal increase in the Fed RRP rate could result in substantial uptake in the RRP program. However, if most money market investors are lending short term at rates substantially higher than the Fed RRP rate (suppose 0.5 percent), then the opposite may be true.
There is some evidence that there is a significant amount of lending taking place only marginally above the effective FFR, as the level of take-up on overnight RRPs with the Fed increases substantially with only very minor increases to the interest rate paid on RRPs relative to the “market repo rate.”42 The market repo rate is the interest rate paid on repo transactions between two private-sector firms. This suggests that, in the above example, only a very small increase in the RRP rate above 1 percent would be required to achieve the desired crowding out effect. In other words, if this holds true, then there is no significant conflict between monetary policy and crowding out.
However, if the Fed sought to crowd out more than $2.6 trillion in short-term debt, there could be a conflict with monetary policy because it would be increasing the total reserves in the system. For example, suppose that the Fed was seeking to reduce outstanding credit in the system (“tightening”) by raising interest rates, while also crowding out private short-term debt by increasing RRP beyond $2.6 trillion. Reserves would have to be increased through asset purchases on the open market, which would increase the reserve balances of banks that sold assets. Holding all else equal, these asset purchases by the Federal Reserve would amount to a higher level of demand for debt instruments at a given interest rate. This higher level of demand would put downward pressure on interest rates, potentially interfering with the monetary policy objective of “tightening” credit conditions by raising interest rates.
If the Fed sought to avoid conflicts with monetary policy, by only using its RRP program to the extent of $2.6 trillion, the remaining $1.7 trillion in short-term debt (we assume total private short-term debt is $4.3 trillion) is still vulnerable to contagion. Thus, while the system would be less vulnerable to contagion ex ante, we will still need to be prepared to deal with it if it still occurs.
But even partially crowding out private-sector issuance raises a number of concerns. If public short-term debt issuance expands during a crisis, as opposed to before a crisis, it may create destabilizing effects for the residual private-sector issuance. Specifically, the expanding public issuance could exacerbate runs by allowing for “disruptive flight-to-quality flows during a period of financial stress and thus could undermine financial stability.”43 It is the rapid change of money market investors from private to public short-term funding that would be destabilizing,44 since the sudden lack of funding sources may leave private institutions incapable of rolling over their debt.45
Further, if the Fed ultimately does succeed in crowding out enough private-sector issuance to appreciably lower contagion risks, the increase in liabilities from newly issued debt will require an increase in assets. In other words, the Fed will have to decide which securities to invest in on the left-hand side of its balance sheet as the liabilities increase on the right-hand side. Should it buy corporate commercial paper or corporate bonds? And if so, from which companies? There would also be the question of what counterparties the Fed would enter into RRPs with on the right-hand side of its balance sheet, raising fairness concerns and concerns about the appropriate role of government in a capitalist system.
In addition there would be the impact on financial institutions that must now replace the short-term debt absorbed by the Fed with longer term debt at higher cost. The impact to those entities no longer able to obtain short-term funding could be particularly damaging because certain valuable activities for the economy should in fact be funded with short-term liabilities. Broker-dealers fund very short-term assets with short-term liabilities, and this makes economic sense. We may make the system safer from contagion but at what cost?
Moreover, if the Fed did expand its balance sheet to appreciably crowd out private issuance,46 the assets it adds would have to bear enough interest to finance payments on RRP. As a result the Fed may have to take on additional balance sheet risk. If the Fed invests in assets that lose value, or if they do not earn sufficient interest, Fed remittances to the Treasury would decrease and taxes or higher deficits would have to make up the difference.
An expanded balance sheet also raises questions about political economy. The average federal funds rate since 1954 is slightly above 5 percent. If the Fed paid this rate on $8.1 trillion of liabilities, this would amount to around $405 billion in interest payments annually. This is comparable to the Department of Defense’s $496 billion budget in 2014.47 Interest payments on this scale may be difficult to defend in the current political climate questioning the role of the Fed in general, and may therefore further jeopardize the Fed’s independence.
My conclusion is that measures to limit short-term funding, particularly through the Fed, deserve further study. The costs of replacing this cheaper short-term funding with more expensive long-term funding, which may not be optimal for particular activities, such as broker-dealer activities, should also be examined. But for now, it would seem highly unlikely that short-term debt issued by the private sector could be sufficiently reduced to obviate the need for a strong lender of last resort and flexible guarantee system.