Bank Reserves vs Reverse Repo

The Federal Reserve's H.4.1 release each Thursday reports the consolidated balance sheet of the Federal Reserve System. On the liability side, Federal Reserve notes (currency in circulation) and Treasury General Account balances are mostly mechanical, while reserves and RRP are the two policy-active liabilities that absorb or release liquidity into the financial system. WRESBAL captures the week-average of bank reserves held at the twelve Federal Reserve Banks, and RRPONTSYD captures the previous-business-day take-up at the Federal Reserve's overnight reverse repo facility. Both are FRED-published, both update at high frequency, and both are direct outputs of the H.4.1.

The arithmetic identity that binds them is the Fed's asset side: as the Fed's Treasury and MBS holdings change through QE, QT, and reinvestment, the corresponding liability change must land somewhere, and most of it lands in either reserves or RRP. When the Fed buys Treasuries from a primary dealer, the dealer's bank credits its account at the Fed (reserves rise) or, if the dealer ultimately delivers cash to a money fund, the money fund's RRP balance rises. The split between the two is determined by the relative attractiveness of the deposit rate (IORB), administered by the Fed for banks, against the RRP rate, administered by the Fed for money funds. Both rates are typically set 5 to 10 basis points apart, with IORB above RRP since June 2021.

RRP take-up peaked at $2.554 trillion on December 30, 2022, the result of a triple-confluence: T-bill supply collapsed during the 2022 debt-ceiling impasse (forcing money funds to park cash at the Fed instead), QT had not yet meaningfully drained reserves, and money fund growth from the rate-hiking cycle pushed an extra $1.5 trillion of cash into government MMFs over 2022. From that peak, RRP drained methodically: $2.0 trillion by August 2024, $1.5 trillion drained over twenty months. The drain ran ahead of QT itself: the Fed's balance sheet contracted by approximately $1.4 trillion from April 2022 through November 2024, but RRP fell by roughly $2.3 trillion over the same window, meaning bank reserves actually rose by about $900 billion as RRP funds rotated back into bank deposits and T-bills.

The Fed slowed QT to a $25 billion monthly Treasury cap in May 2024, slowed it further to $5 billion in February 2025, and ended balance-sheet runoff entirely in December 2025. The reasoning was reserve adequacy: with RRP near zero, additional QT would directly drain reserves rather than RRP, and the SOFR-IORB spread had begun to widen in late 2024, signaling that the banking system was approaching the bottom of the 'ample reserves' band. WRESBAL has stabilized in a $2.85 to $2.95 trillion range through the first four months of 2026, which the New York Fed's Markets Group has explicitly described as 'modestly above' the lower bound of ample reserves.

The Bank Term Funding Program (BTFP), announced March 12, 2023 in response to the Silicon Valley Bank failure, lent against par-value Treasury and agency MBS collateral and provided the banking system with a non-discount-window liquidity backstop. BTFP balances peaked at $167.4 billion in January 2024 before the program closed to new loans on March 11, 2024, and the outstanding balance has since been amortizing as one-year loans matured. The BTFP balance shows up on the asset side of the Fed balance sheet as 'Loans' rather than as Treasury or MBS holdings, but it produced the same liability-side effect: reserves rose as banks pledged collateral and received cash. The program structure also created a regulatory arbitrage where banks could pledge depreciated long-duration securities at par, capturing a balance-sheet benefit that exceeded the program's stated stabilization purpose.

The practical implication is that the WRESBAL series during 2023-2024 was elevated by an artificial $50 to $167 billion BTFP overlay that was not part of the underlying reserves-versus-RRP relationship. The BTFP overlay is now mostly amortized, with outstanding balances below $5 billion as of April 2026, but the legacy effect is that the 2023-2024 WRESBAL trajectory was misleading as a measure of 'natural' reserve demand. Disciplined H.4.1 readers subtract BTFP loans from headline WRESBAL when comparing across periods, and the post-March 2024 series now reflects the natural reserves-versus-RRP balance without the program overlay. The Federal Reserve's 2024 Annual Report explicitly acknowledged the WRESBAL distortion in its discussion of post-March 2023 banking-system liquidity dynamics.

The Convex Net Liquidity Impulse (CNLI) decomposes Fed liquidity into three components: outright Federal Reserve assets, the Treasury General Account, and RRP balances. The current configuration, with WRESBAL near $2.9 trillion and RRP near $300 billion, registers as 'reserves-funded ample' rather than 'RRP-buffered ample': the Fed has effectively transitioned from a regime where money funds held the marginal liquidity dollar at the RRP facility to a regime where banks hold the marginal liquidity dollar as reserves. The transition is mechanically meaningful because RRP balances do not feed into private credit creation while reserves do, which is why the post-2024 transition has supported moderately easier financial conditions despite the Fed's restrictive policy stance. The Federal Reserve Bank of New York's 2024 Markets Group review estimated the implicit easing from the regime shift at approximately 25 basis points of equivalent rate-policy stance.

The disagreement watch is on RRP rebuilding rather than on further drain. RRP take-up has shown brief spikes above $300 billion at quarter-end dates (March 31, 2026 closed at $312 billion as money funds parked excess cash before the quarter-end SLR window) but has not sustained any move above that level. The CNLI dashboard treats sustained RRP rebuilding above $500 billion as a signal that money funds are de-risking from T-bills and bank deposits, which historically (December 2018, March 2020, October 2022) has marked the early innings of broader risk-off positioning. The April 2026 readings sit well below that threshold.

The pair's signal is most actionable when one leg moves materially in the absence of a corresponding move in the other. Three specific configurations have produced consistent outperformance in the post-2009 record. First, RRP rising while reserves fall by an offsetting amount, with no change in the Fed asset side, signals that money funds are de-risking from bank deposits into Fed-administered RRP; this is the late-2007 template (RRP did not exist then, but the equivalent flow was bank deposits to Treasury bills) and the March 2020 template (RRP rose by $200 billion in three weeks). Second, RRP falling while reserves rise by an offsetting amount, with no change in the Fed asset side, signals that money funds are reaching for yield in T-bills and bank deposits; this is the 2023-2024 template that produced the $2.3 trillion RRP drain and supported the equity rally. Third, both legs falling together while the Fed balance sheet contracts is straightforward QT-on-reserves, which is what would re-emerge if the Fed restarted balance-sheet runoff.

The April 2026 configuration is the post-transition steady state: RRP near $300 billion, reserves near $2.9 trillion, balance sheet flat at $6.7 trillion. The pair does not currently provide a directional signal but does provide regime context: the next meaningful move will be either RRP rebuilding above $500 billion (de-risking signal) or reserves falling below $2.7 trillion (reserve scarcity signal). Both would represent material regime shifts and would require macro overlays to update broader liquidity assumptions accordingly. The historical base rate for either threshold breach producing a 30-day move in the broader S&P 500 of more than 5 percent is 67 percent in the post-2009 record, with the direction depending on which threshold is crossed first.

Three specific events would break the current reserves-versus-RRP equilibrium and produce signals that the pair currently is not generating. First, a debt-ceiling impasse that forces the Treasury to draw down the TGA below $200 billion: the resulting reserve buildup would be mechanical (TGA-to-reserves substitution) and would not represent any underlying liquidity easing, but it would produce a temporary downward pressure on short rates that the unfiltered reader would misinterpret as a Fed pivot. The 2023 debt-ceiling impasse produced exactly this dynamic from January through June 2023, with reserves swelling by approximately $400 billion before normalizing after the June 2023 resolution.

Second, a regulatory-driven shift in money fund allocation: the SEC's 2023 reform of institutional prime money market funds moved approximately $400 billion from prime to government MMFs over 2024, which mechanically shifted RRP eligibility without representing any change in cash-versus-equity allocation. Third, a renewed banking stress event: a regional bank failure on the scale of Silicon Valley Bank or a stress event on a globally systemic bank would produce a flight from bank deposits to MMFs and from MMFs to RRP, with reserves falling and RRP rising in a discrete step. The 2023 episode produced approximately $400 billion of such reallocation over a four-week window. Watching the daily SOFR-IORB spread, the H.8 deposit data, and the weekly H.4.1 print together is the cleanest cross-check on whether the current ample-reserves equilibrium is being challenged.