By now, anyone keeping an eye on financial news will be aware that some sort of crisis took place in the repo market in late September. Although the general outline of the drama may be familiar to some of my readers, I suspect that the specifics remain unclear for most… and for good reasons. Making sense of what happened requires a clear understanding of several complex topics ranging from the functioning of the Federal Reserve to the way financial institutions finance their day-to-day operations. Why write about the September turmoil now that it seems to be over? Simply put, because what took place is likely to have broad effects, notably on the future actions of the Fed. More on that later. Where to begin? Explaining what a repo is seems to be a good starting point (bear with us). A repo – repurchase agreement – is a way for a financial institution to borrow or lend cash with financial assets as collateral. In a typical repo transaction – a secured overnight financing operation – one party sells a security, like a Treasury bond, to another for cash and agrees to repurchase the security at a slight premium at a prescribed time, often the next day. The repo rate determines the size of the premium the borrower will pay to repurchase the security. Such transactions underpin much of the short-term borrowing and lending that financial institutions carry out to manage their daily cash flow and are thus vital to the proper functioning of the financial system. Trillions of dollars are exchanged every day in the U.S. repo market. Most of the time the repo market operates without glitches. But in the week of September 16 it went berserk. The repo rate, which had been stable around 2%-2.5%, suddenly spiked to 10%, leaving those needing to borrow cash exposed. On the 17th, after some hesitation, the New York Fed offered $75 billion in alternative funding, effectively taking over from the private institutions that normally participate in that market. The intervention was repeated daily (and enhanced) in an effort to keep funding cheap and prevent contagion to other areas of finance. Repo rates eventually fell back to more normal levels, but a certain nervousness remained. After all, it was the first time the Fed had to intervene to stabilize the market since the crisis of 2008-09. Why did it happen? The first explanations to come out in the media emphasized a surge in demand for cash in the week of September 16. Like much else in economics, repo rates are governed by supply and demand. When the volume of liquidities sought by would-be takers increases faster than supply, liquidity become dearer. There were indeed specific circumstances fueling demand for cash in mid-September. First, companies in the U.S. had to pay tax instalments. At the risk of sounding a bit patronizing, let me explain how tax payments work from an accounting perspective. When a company writes a cheque to the government, the money moves from its bank account to the public treasury. Concretely, the disbursement reduces the bank’s reserves held at the Fed and increases the balance in the Treasury’s central bank account by the same amount. The disappearance of the tax-payment amount from the bank’s reserves leaves it with less cash on its books. All else equal, this will reduce this bank’s propensity to lend cash on the repo market. It may even look to borrow cash itself to restore its pre-disbursement liquidity ratios. Compounding the effects of tax payments, the U.S. Treasury issued a large volume of debt in the days leading up to the repo market crisis, $77 billion of which settled on September 16. In a process similar to tax payments, when a dealer buys a debt auction, a transfer of reserves to the Treasury’s account occurs leaving less funds available to lend. With the U.S. budget deficit swelling at the moment, such large auctions are bound to become more frequent. This could increase the risk of disruption on the repo market.


To summarize, repo rates spiked because of a surge in demand for cash. The latter was caused by an abnormal depletion of reserves, itself the combined effect of tax payments and large Treasury auctions. But tax payments and Treasury auctions have happened before. Why did the wheels come off this time? Because reserves are less abundant than in the past. Explaining why requires delving a little more deeply into central bank accounting. The Fed’s assets (mostly Treasuries and mortgage-backed securities) must, like those of any other bank, be equal to its liabilities (currency in circulation and reserves, among other items). In response to the crisis of 2008- 09, the Fed boosted the size of its balance sheet from about $870 billion to roughly $4.5 trillion. Most of the increase came from aggressive bond-buying by the central bank, known as quantitative easing or QE. As the Fed bought bonds in the secondary market, it paid for them by creating reserves that it deposited to the account of the selling bank. As a result of QE, the amount of outstanding reserves surged from virtually nothing to a peak of $2.7 trillion in 2014. Then, as the Fed began to reduce the size of its balance sheet, the process reversed and reserves started shrinking. Bank deposits were hit disproportionately, since other liabilities – mainly currency in circulation – continued to expand. Since 2014, reserves have shrunk roughly 50% to $1.3 trillion, whereas the Fed’s balance sheet is down just 15%.


Let me be clear here: the reserves held at the Fed are still abundant. U.S. financial institutions are required to keep prescribed minimum reserves at the Fed (currently less than $200 billion) to cover deposit liabilities. These are called “required” reserves. Total reserves at present exceed that amount by more than $1 trillion. Clearly, recent tax payments and Treasury issuance did not threaten to reduce reserves below the required minimum. So if financial institutions did not risk crossing the regulatory threshold, why were they reluctant to lend their reserve cash through the repo market? The answer to that question is closely linked to new post-crisis regulations and changing bank preferences. Since the recession, regulators have demanded that systemically important financial institutions hold minimum levels of high-quality liquid assets (HQLA) to prevent the kind of acute liquidity shortages observed in 2007-08. Such assets include deposits at the Fed, Treasury bonds, mortgage-backed securities and others. Though reserves and Treasury bonds are treated equally in the overall liquidity ratio, banks have shown an increased preference for the former. There are many reasons for this inclination. First, apart from liquidity coverage ratios, regulators track what is called “intraday liquidity” – cash that can be immediately accessed by financial institutions. That Treasuries cannot be included in intraday liquidity – Treasury transactions settle on day T+1 – effectively puts a premium on reserves. Second, since the crisis, reserves have become more attractive relative to other liquid assets. Before 2007-08, reserves held at the Fed earned zero interest, putting them at a significant disadvantage to, say, Treasuries. As a result, most financial institutions opted to keep their average reserve as close as possible to the minimum regulatory requirement. Today, the Fed pays interest on reserves (Interest Rate on Excess Reserves or IOER), greatly reducing the opportunity cost of holding them. The final reason banks now want to hold more reserves has to do with liquidity risk, as was clearly explained in April by New York Fed vice-president Lorie K. Logan 1: “There are new factors driving reserve demand, and that demand is, and will likely continue to be, much higher than it was before the crisis. To start, the crisis appropriately changed attitudes toward risk and increased focus on managing liquidity risk. The subsequent creation of a liquidity regulatory framework for larger banks reinforced the benefits of holding unencumbered liquid assets for both domestic and foreign banks. While these liquidity regulations do not themselves impose any specific requirements to hold central bank reserves, many banks now use internal models that also estimate the very short-term liquidity they need to hold in order to be prepared for a stress scenario. These models make assumptions about a bank’s reduced access to funding and ability to sell securities during earlier stages of a stressed period. The bank may therefore decide to hold a portion of its liquidity buffer as reserves on an ongoing basis as a precautionary measure.” In other words, since Treasuries carry an interest-rate risk in the internal stress-test scenarios of financial institutions, they are not treated on a par with reserves held at the Fed. All of these factors mean that financial institutions now want to hold a lot more reserves than they did in the past. But exactly how much is not clear. In an effort to determine the lowest comfortable level that would be adequate in the system, the Federal Reserve ran surveys of financial institutions. The latest one found that participants would be comfortable with a level of reserves ranging from $800 to $900 billion. These results left the Fed confident that current reserves were unlikely to trigger the kind of crunch that occurred in the repo market. But the Fed was wrong. Surveys are estimates that can change over time for reasons specific to each institution or because of change in market environment (Logan 2019). For instance, it is possible that all the recent talk of the possibility of recession increased the level of reserves that banks wanted to have. It could be also that total demand for reserves is higher than the sum of each bank’s requirements in an environment where reserves are not distributed efficiently. Each institution determines what mix of HQLA is right for its need, and some may want to hold a lot of reserves – much more than they are required to – and still be unwilling to lend on the repo market. The upshot is that nobody is quite sure what would be an appropriate level of reserves in the system, i.e. a level that would prevent the kind of chaotic movements that took place on the repo market in September. So how does the Fed make sure that recent events don’t recur periodically?


In the near term, the Fed is likely to continue providing ad hoc short-term funding to keep overnight rates under control; latest communications from the central bank stated it would do so through November 4. But a longer-term fix is needed. At the moment two possibilities are being discussed. The first is for the Fed to resume expansion of its balance sheet. As previously noted, reserves grow when the Fed purchases assets, so buying Treasuries sounds like the obvious way to make reserves swell. What amount are we talking about here? Well, since demand for money in circulation is expected to keep growing (“organic” growth), the Fed will need to expand its balance sheet by approximately $75-100 billion a year just to keep up with that demand. But in light of recent market action, it appears that the current supply of reserves is insufficient. The Fed will thus need to be more aggressive in its asset purchases. Recent estimates suggest that an additional $200-$500 billion in total would be needed to build a sufficient reserves buffer.


Currently this crisis oes not affect you directly but if it happens again and for a long period the ripple effect can cause unwinds in leveraged positions and credit portfolios resulting is wider credit spreads and possibly lower equity prices.