All other things being equal, the public spending achieved by drawing on the TGA will automatically lead to an increase in bank reserves with the Fed (Box 3). However, this positive impact could be partially, if not entirely, offset by money market fund arbitrage. With T-bill issues becoming scarcer, money market funds are likely to increase their "deposits" with the Fed via the ON RRP facility (which will automatically destroy part of the banks' reserves, see Box 2). Between 19 February and 5 March, the US Treasury withdrew USD 216 billion from its account. However, bank reserves only increased by USD 105 billion over the same period, due, on the one hand, to the continuation of QT2 (Fed assets down by USD 26 billion) and, on the other, to an increase in MMF "deposits" (USD 70 billion), and an increase in other Fed liabilities (USD 15 billion). If loans on the private repo markets remain more attractive than the ON RRP facility (as is currently the case, see Chart 4), the effect of the reduction in the TGA on reserves will be very positive and will cushion the impact of QT2 for a few months. Otherwise, there will be little or no effect.
In any case, the overall pool of central bank liquidity (bank reserves and MMF deposits with the Fed) will increase during the debtceiling negotiation phase. Later, a political compromise will enable the US Treasury to replenish its account with the Fed by issuing securities. The low attractiveness of the ON RRP facility will encourage MMFs to leave it and invest in T-bills. In the event of a partial replenishment of US Treasury assets, the net effect on reserves of this episode (excluding the QT effect) could be positive. However, it is likely that the US Treasury will have to replenish them in full (especially if the new administration estimates that its cash requirements should there be a cyber attack are at the same level as the previous administration), and that the net effect will be zero. Whatever the net effect, for a few months, the indicators monitored by the Fed (see below) are likely to become less relevant. In August 2019, QT1 was ended. At the same time, the ceiling on federal debt was suspended, allowing the US Treasury to replenish its holdings with the Fed. The amount of central bank money proved insufficient just one month later (Chart 5).
IN 2019, THE REPO MARKET CRISIS OCCURRED ONE MONTH AFTER THE SUSPENSION OF THE FEDERAL DEBT CEILINGThe difficulty of assessing banks' reserve requirements
As it does not know precisely what the optimum amount of reserves should be (neither too little nor too much, but sufficiently "ample" to avoid any risk of stress that would require it to inject central bank money as a matter of urgency), the Fed has set itself the objective, since the launch of QT2, of keeping its reserve supply above a floor, close to the level at the time of the crash in 2019. In May 2024, it presented two possible trajectories for the evolution of its balance sheet[7]. They both assume that the reduction in the balance sheet will end in 2025 and that, after a one-year pause[8], asset purchases will resume in order to preserve the stock of reserves at 8% (low assumption) or 10% (high assumption) of GDP. On the basis of reserve demand curve simulations, Afonso, Giannone, La Spada and Williams (2022)[9] also estimated that a reserves-to-bank-assets ratio of less than 11% would suggest a scarcity of reserves, a ratio of more than 11% but less than 12-13% would indicate a sufficiently "ample" stock of reserves, and above 13%, an “abundant” stock.
Banks' outstanding reserves with the Fed averaged around USD 3,200 billion in the first two months of 2025, equivalent to 11.3% of GDP and 14% of banking assets (compared with USD 1,400 billion, equivalent to 8% of GDP and the same amount of banking assets, at the time of the September 2019 repo market crisis). According to the parameters adopted by the Fed, it would therefore still have room for manoeuvre to reduce the size of its balance sheet. By mid-February, however, the stock of reserves in excess of the threshold that marks the borderline between sufficient and insufficient reserves stood at just USD 370 billion. Assuming that the current pace of QT2 is maintained, the Fed’s securities portfolio should shrink by a further USD 400 billion by the end of the year. All other things being equal, and based on our GDP growth forecasts, QT2 could be completed by summer 2025.
However, there is a risk that these floors have lost their relevance. For various reasons, banks’ reserve requirements are changing (Box 1). The difficulty lies in assessing to what extent.
In the United States, the minimum level of reserves desired by the large-scale banks is all the more important given that their liquidity constraints are particularly stringent (see below), and the stigma associated with the Fed's lending windows deprives them of access to central bank money when needed. Various solutions are being considered to correct the very negative perception of this issue by banking supervisors and bank managers since 2008. The main breakthrough in this area since the March 2023 bank run is the clarification provided by the Fed in its publication of “frequently asked questions” last summer[10]. In this publication, it states that, as part of their internal liquidity stress tests (ILST), banks may consider replenishing their stock of reserves by mobilising their assets pre-positioned with the Fed (via the discount window or the Fed's reverse repurchase facility) or with Federal Home Loan Banks[11] (via FHLB-secured loans), provided that they mobilise "highly liquid" assets (identical to the securities eligible for high-quality liquid assets in the LCR numerator)[12].
This easing aims to offset the stigmatising effect - in the eyes of bank supervisors and managers - of using the Fed’s emergency facilities. It must convince them of their usefulness, within the regulatory framework, in meeting banks' immediate liquidity needs (and encourage banks to prepare to make use of them by pre-positioning assets). By giving US Treasury securities and mortgage-backed securities (MBS) issued by mortgage guarantee and refinancing agencies the status of quasi-substitutes for central bank reserves, within the framework of ILST (which is the most restrictive liquidity requirement in the United States), this easing could also reduce the desired minimum level of reserves. However, it risks crowding out the holding of less liquid or illiquid assets (such as loans to the economy) in favour of liquid assets (Treasuries, MBS) and diverting the discount window from its primary purpose (transforming illiquid assets into liquidity for solvent banks faced with a liquidity shock)[13]. In addition, its ability to reduce the stigma attached to the Fed's lending facilities has yet to be demonstrated.
A list of leading indicators to be added
The Fed has added to its range of tools for detecting potential reserve shortages. In particular, since October 2024, it has published a “real-time” estimate of the elasticity of the effective federal funds rate to a change in the aggregate stock of reserves (Reserve Demand Elasticity, RDE)[14]. According to the Fed, this estimate would make it possible to distinguish between periods when reserves are “abundant” and those when they are “ample” or even scarce[15]. At the beginning of February, the RDE remained close to zero.
Other indicators are also monitored[16], such as the proportion of interbank payments settled at the end of the day (after 5 p.m.)[17], the size of banks’ daylight overdrafts with the Fed[18], the volume of borrowings by US banks on the federal funds market[19] and the share of repurchase agreements charged at or above the interest rate on reserves[20].
Based on these indicators, the Fed considers that reserves remain abundant for the time being (Logan, 2024 and Perli, 2024 and 2025)[21]. In fact, the pressure seen on repo market rates in the second half of 2024 (Chart 6)[22] was largely caused by specific calendar events (when financial accounts are closed at the end of the quarter and at the end of the year, financial intermediaries are encouraged to reduce their balance sheet exposures, in particular by not renewing their repo loans; on the settlement dates for issues of US Treasury securities or tax payments, the stock of reserves is automatically eroded). Against a backdrop of shrinking global liquidity, the Fed does not consider these tensions to be alarming, especially as they were modest and quickly dissipated (Gowen, Perli, Remache and Riordan, 2025)[23].
THE REDUCTION IN THE GLOBAL SUPPLY OF CENTRAL BANK MONEY IS BEGINNING TO BE FELT ON THE MONEY MARKETSThe indicators monitored by the Fed nevertheless need to be supplemented. In particular, the Fed seems to be underestimating the effects of the growth in centrally cleared repurchase agreements for Treasuries, which reduces quarter-end tensions but increases banks' reserve needs (see below). In addition, while the average value of banks’ daylight overdrafts with the Fed remains low, a sign of the abundance of reserves at the aggregate level, the peaks in daylight overdrafts widened significantly in the second half of 2024, returning to their 2019 level[24]. Finally, the Fed’s focus on the federal funds market raises a few questions. This is because 1) daily outstanding loans are very modest (an average of around one hundred billion dollars lent each day, compared with at least[25] USD 4,000 billion on the Treasuries repurchase agreement markets), and 2) the number of participants is normally limited for regulatory reasons (the GSEs make up most of the loans and the US branches of foreign banks make up most of the borrowings[26]).
Admittedly, in 2018 and 2022, as liquidity needs became more pressing, the US regional banks made greater use of the federal funds market. However, the recommendations published in August 2024 by the Fed and the FDIC on the resolution plans imposed on category 2 and 3 banks[27] could reduce their interest in this market. The supervisors are not advocating a framework as restrictive as that applied to systemic banks, but are recommending daylight monitoring of liquidity risks. However, borrowing federal funds is not suitable for meeting this type of requirement. On the federal funds market, loans are overnight. Borrowed funds are generally repaid in the morning (around 5.30 a.m. to 6 a.m. New York time) and borrowings renewed at midday. The banks are therefore left with no liquidity for a few hours. Borrowing federal funds enables banks to obtain central bank money to settle an interbank debt, cover a payment delay or meet the Basel LCR liquidity constraint, but it is not suitable for meeting the liquidity requirements specific to the resolution plans.
On the other hand, FHLB deposits with banks are relevant leading indicators, which would be a welcome addition to the list of those already monitored by the Fed. The speed of their expansion, first in 2018 and then during 2022, on the eve of two episodes of extreme liquidity tension (September 2019 and March 2023), and the remuneration offered in return (which may have exceeded that of the banks' reserves with the Fed) suggest that, unlike borrowing federal funds, they improve the daylight and daily liquidity positions of the large-scale banks (Charts 7 and 8). Their supervision seems all the more opportune as, since last January, the exposure limits per counterparty, applied to the deposit accounts of the FHLBs, have been raised to the same level as those set for their federal funds loans[28].