“[T]he Committee reaffirms its intention to implement monetary policy in a regime in which an ample supply of reserves ensures that control over the federal funds rate and other short-term rates is exercised primarily through the setting of the Federal Reserve’s administered rates, and in which active management of the supply of reserves is not required.”
FOMC Statement Regarding Monetary Policy Implementation, October 11, 2019.
“This is not QE. In no sense is this QE.”
Federal Reserve Chair Jerome Powell, cited in The Wall Street Journal, October 9, 2019.
On September 17, the overnight Treasury repurchase agreement (repo) rate spiked to 6%—up from just 2.2% a week earlier and the highest level in more than 15 years (see DTCC GCF repo index). The repo market—a key source of short-term wholesale funding for U.S. intermediaries—became so disorderly that some trades reportedly occurred at 10%. Similarly, in the broader market for secured overnight financing, the top 1 percentile of trades occurred at a rate of 9%, far above the weighted average of 5.25% (see chart). Even the federal funds rate rose a bit above its then prevailing target range of 2 to 2¼ percent. Oddly, all this turmoil occurred at a time when the Fed had begun lowering its policy rate for the first time in more than a decade and market participants anticipated further policy easing ahead.
Secured overnight financing rate (SOFR), April 2018-25 October 2019
Note: The solid black line shows the SOFR rate, which is a weighted average of all trades. The area shaded in pink between the dotted lines shows the range between the 1st percentile and 99th percentile of trades. Source: FRBNY SOFR data.
What led to this sudden disruption in short-term funding markets that had been relatively calm in recent years? Had the Fed lost control? The financial press was full of speculation. Many people noted upcoming large quarterly corporate tax payments as a source of heightened demand for funds. Others suggested that liquidity regulations were making banks’ demand for reserves overly rigid, so that they would be unwilling to lend in the repo market, even when it appeared profitable.
In our view, the explanation for the sudden rise in overnight interest rates is straightforward: the shrinkage of the Federal Reserve’s balance sheet that began in October 2017 reduced the aggregate supply of reserves from a level like RS0 (in the chart below) gradually to one like RS3—where banks’ demand for reserves was insensitive to interest rates. Put as simply as possible, the reduced supply hit a point on the demand curve for reserves that was steeply upward sloping—or highly inelastic—consistent with a point like D in the chart. Consequently, large temporary fluctuations in the supply of reserves that would have had virtually no impact even a few months ago, triggered sizable upward interest rate fluctuations. (Note that another Federal Reserve policy tool—the overnight reverse repo (ON RRP) rate—limits downward fluctuations of market interest rates.)
The Market for Reserves
Notes: The red curve labeled RD is the demand for reserves. The vertical blue lines depict the evolving supply of reserves (RS) that gradually shifted left as the Fed’s balance sheet shrank. The gray-shaded area shows the target range for the federal funds rate that prevails when the effective federal funds rate is at point D.
Consistent with this view, the Federal Reserve recently took action to prevent a recurrence of the September disorder (see opening FOMC citation). After weeks of ad hoc daily short-term repo operations—adding on average about $150 billion to its assets in the month after September 17—at an unscheduled video conference meeting on October 11, the FOMC agreed to additional regular purchases of Treasury bills at least into the second quarter of 2020. The goal of this balance sheet expansion is to maintain reserve balances at least as high as their level in early-September before the turmoil began (a level like RS2). To implement this new directive, the FRBNY Trading Desk announced plans to purchase Treasury bills starting in mid-October—initially at a pace of $60 billion per month. These outright purchases will partly substitute for (and potentially displace) the massive daily operations that began to expand the Fed’s overall liabilities last month. While they are helping to restore market order, these actions leave open the question—which we previously addressed in some detail—whether the Fed should eventually create a new standing repo facility that would satisfy sudden spikes in demand for non-reserve Fed liabilities.
In the remainder of this post, we discuss the evolution of the supply and demand for reserves in recent years. We argue that, because no one—including the Fed—knew the precise level of reserves at which the demand curve would become inelastic (like point D in the chart above), an episode like the one on September 17 was virtually inescapable. If our diagnosis of the cause is correct, then recent actions should help put the issue to rest. Yet, given the inevitability of the event―that the day would come when reserve supply hit the inelastic part of the reserve demand curve―the Fed could (and should) have been prepared. If so, it could have avoided even a temporary dent in its well-deserved reputation for operational prowess.
Let’s start with the basics. In its monetary policy operations, the Fed aims to control short-term market interest rates like the federal funds rate or SOFR. If the Fed wishes to do so using its administered rates—like the interest rate on excess reserves (IOER) or the ON RRP rate)—then it must supply reserves well in excess of demand. Reserves are the deposits that banks hold at the central bank, and the Fed determines their aggregate volume. Such an “ample” supply of reserves (shown in blue in the next chart) is precisely what the Fed engineered with the massive expansion of its balance sheet during and after the financial crisis of 2007 to 2009. On the chart above, the regime of ample reserves corresponds to the supply curve labeled RS0, which intersects the demand curve where it is almost perfectly flat (point A) and keeps the market interest rate close to the IOER (see here).
Federal Reserve liabilities (Billions of dollars, Wednesdays), 2008-October 23, 2019
Source: FRED (Federal Reserve H.4.1).
Yet, as soon as the Fed’s aggregate liabilities stopped rising in 2014, reserve supply peaked (at $2.8 trillion) and began to decline. Looking at the chart above, over the next three years, with the overall size of the balance sheet roughly unchanged at $4.5 trillion, the ongoing rise of non-reserve liabilities gradually displaced reserves. Currency (shown in gray) grows smoothly over time and is the most important of these other liabilities. Over the past decade, currency has expanded at a 6.8% annual rate—even faster than the economy. Assuming that this continues, over the next year, additions to currency will take up a further $10 billion of Fed liabilities each month. So, unless the Fed purchases $10 billion per month in securities, reserves will fall. (We should note that most of this additional currency is in the form of $100 bills, much of which finances untoward activities: see our previous discussion here.)
From October 2017 until July 2019, the Federal Reserve also shrank its overall balance sheet by allowing some of its maturing assets to expire without replacement. Since both sides of the balance sheet must be equal, the decline in assets was matched by a decline of liabilities. With currency continuing to grow, reserve supply declined markedly further. As the chart above shows, reserves accounted for more than all of the $700-billion decline in Federal Reserve liabilities, sinking by $724 billion to $1.5 trillion by August 2019.
When the Fed began reducing the size of its balance sheet, reserves were so ample that shocks to other liabilities that resulted in sudden reserve declines of $100 billion or more—say, as a result of swings in the Treasury’s account at the Fed (in yellow above)—had virtually no impact on the market interest rates. Put differently, with reserve supply near $2 trillion, bank demand was almost perfectly elastic. Graphically, shifts in supply between RS0 and RS1 in the second chart would result in moves between points A and B along the flat portion of the demand curve. However, as reserves continued to decline, the intersection of supply and demand shifted gradually left to a part of the demand curve that had a mild upward slope. So, a supply shift from RS1 to RS2 produces a shift from point B to point C, with the interest rate rising accordingly. In an earlier post, we described how, beginning around March 2018, the effective federal funds rate began to rise very slightly relative to the interest rate on excess reserves (IOER), and eventually surpassed it earlier this year, suggesting that the reserve supply had declined to the point where reserve demand was gently upward sloping.
The central bank’s operational challenge is to predict where the reserve demand curve becomes sharply upward sloping even though, prior to September 17, no one had observed such interest rate insensitivity of reserve demand for more than a decade. The Fed has several ways to anticipate the level of reserve demand, but they proved imperfect. The most important is that they ask market participants. For example, using surveys of senior bank financial officers, the Fed sought to identify the minimal level of reserves that the banking system needed to be “comfortable.” The latest survey—conducted in August and released on October 18—suggested that banks would be satisfied with a reserve supply more than 40 percent below levels prevailing at the time. That likely gave officials confidence that the Fed was still operating along the flat portion of the reserve demand curve. However, the same survey showed that banks’ holdings of reserves in excess of their minimal comfort level reflected their desire for a “cushion” against potential outflows. Presumably, that is precisely the kind of internal liquidity risk management that regulators wish to see and encourage. But it means that the surveys were less useful as a barometer of precautionary reserve demand, which is what matters.
In addition to periodic surveys, the Trading Desk of the Federal Reserve Bank of New York conducts daily surveillance of active dealers in the money markets to anticipate any dislocations. That is, they speak to people each morning. Late in the week before September 17, money market rates were already edging higher, presumably out of concern about upcoming drains on liquidity (a corporate tax payment deadline and new Treasury issues to absorb). And, the disorder in the overnight secured financing market was readily apparent on September 16, with the 99th percentile rate (4.60%) jumping more than 2 percent above an elevated market rate (2.43%).
Our point is not that the Fed should have anticipated that the disorder was going to occur on exactly September 17. It is that the central bank was operating in a way that made an eventual disruption inevitable, allowing reserve supply to shrink until eventually it would intersect the demand curve in a steep upward-sloping region. To put it slightly differently, the Fed sought to operate with a minimal buffer of ample reserves, rather than with a truly abundant amount (see, for example, here). Knowing that the day of reckoning would come, the Fed could have had a plan on the shelf that it could activate immediately, minimizing the length of the disruption. Furthermore, had officials crafted and announced the details of this shelf plan in advance, they could very well have reduced the extent of the disorder. They might also have short-circuited public speculation—such as about why banks may have been reluctant to lend in the repo market on September 17 (see, for example, here)—that was largely beside the point. As we have seen, simply restoring reserves to the early-September level calmed the money markets.
Importantly, our analysis supports the Fed’s argument that the current balance sheet expansion involves virtually no monetary policy stimulus (see the opening citation from Fed Chairman Powell). To be sure, we define quantitative easing (QE) as an increase in the supply of reserves beyond the minimum needed to meet the interest rate target. So far, however, the expansion of the balance sheet has merely stabilized the supply of reserves, while accommodating increased demand for non-reserve Fed liabilities. And, since the Fed plans to invest the proceeds from its balance sheet expansion in Treasury bills, even if it increases the supply of reserves to restore a healthy buffer, it is engaging in virtually no transformation of liquidity, maturity or credit. Viewing the federal government’s balance sheet on a consolidated basis, all the Fed is doing is swapping one very short-term government liability for another.
To conclude, it is important to keep in mind that if the Fed wishes to maintain an “ample” buffer of excess reserves, then it will have to continue expanding its liabilities to accommodate rising currency demand. Normal growth of the banking system should be accompanied by some growth in reserves as well as in currency. The Fed also will have to ensure that the buffer is large enough to weather temporary changes in non-reserve liabilities. For example, since the central bank began shrinking its balance sheet, the standard deviation of weekly swings in the Treasury account at the Fed has been $41 billion, with the maximum change at $124 billion. And, while there may be no shortage of new Treasury debt overall, very short-term bills already enjoy a premium (see, for example, Carlson et al and Greenwood, Hanson and Stein). Accordingly, given their stated plan to buy large volumes of Treasury bills, we hope that the Fed seeks a new “accord” with Treasury debt managers to ensure an adequate supply consistent with effective monetary policy operations.
Acknowledgement: Without implicating him, we thank our friend and Stern colleague, Richard Berner, for very helpful suggestions.