Reverse Repo vs S&P 500

The Fed Overnight Reverse Repo facility (ON RRP) was activated as a daily auction in September 2013 to put a floor under money-market rates. It sat near zero through 2017-2020, then exploded from $10 billion at year-end 2020 to $2.554 trillion at the December 30, 2022 peak (a year-end surge driven by usual quarter-end pressures plus the fed-funds-target lift to 4.25-4.50 percent). The drain from peak to near-zero has been one of the most-watched liquidity sequences of the 2020s, because the policy-watch thesis (articulated by Goldman's Marquee, JPMorgan's Treasury Markets and Federal Reserve Bank of Kansas City research) was that RRP-sourced liquidity was the buffer absorbing quantitative tightening before bank reserves had to drain. The pair tests an institutional thesis published most cleanly by the Federal Reserve Bank of Kansas City in November 2023 and updated through 2025: as long as RRP balances were available, QT could continue without stressing reserves; once RRP drained to under $200 billion, any additional QT would press directly on reserves. The April 2026 reading of $0.7 billion (effectively zero, with quarter-end blips) is the live test of that thesis, and the SPY advance through the entire drain provides the empirical answer that the thesis was largely correct: the buffer absorbed the QT without producing the equity weakness simpler liquidity models had predicted. Reading the pair in isolation produces the simplest and most misleading version of the question; reading it inside a multi-input net-liquidity composite is the institutionally rigorous frame, and is what makes the pair operationally useful rather than a frequently-cited but rarely-actionable indicator.

December 30, 2022: RRP $2.554 trillion peak. SPY $381.34 close. The peak coincided with the worst SPY year since 2008 (negative 18 percent total return for 2022). Mid-2023: RRP $1.95 trillion average. SPY rallied 26 percent over H1 2023 despite the regional banking stress (March 2023 SVB and Signature failures), refuting the simple thesis that draining liquidity would hold equities back. End 2023: RRP $700 billion. SPY $476.51, up 24.7 percent for the year. Mid-2024: RRP $400 billion (the Wolf Street widely-cited reading), SPY 23 percent ahead year-to-date. December 2024: Fed trimmed the ON RRP rate by 5 bps on top of the rate cut on December 18, 2024, sending RRP to $98 billion within a week, the lowest since April 2021. April 29, 2026: RRP at $0.7 billion, SPY at $568.40 (up 49.1 percent from the December 30, 2022 close). The empirical record is decisive: RRP drained from $2.55 trillion to near-zero, a $2.55 trillion liquidity withdrawal, and SPY advanced 49.1 percent over the same window. The simplest version of the QT-as-equity-headwind thesis is therefore falsified at the multi-year horizon, and any future analysis of the period needs to account for the structural reasons the drain did not produce the predicted equity weakness. The 28-month decline of 99.97 percent of facility balance ranks as the longest sustained liquidity withdrawal of the post-2008 era, dwarfing the 2017-2019 QT episode in both pace and magnitude, and producing the cleanest empirical record available for testing liquidity-versus-equity theses.

The RRP drain did not produce equity outflows; it produced reallocations from money funds into Treasury bills. Three transmission channels dominated: (1) Treasury bill issuance accelerated after the June 2023 debt-ceiling resolution, with bill share of total Treasury issuance rising from 16 percent to 22 percent over 2023-2024, providing the alternative high-quality short-duration asset that money funds rotated into; (2) money-fund AUM grew from $4.7 trillion at end-2022 to $6.8 trillion by April 2026, indicating that the cash buffer was not deployed into risk assets but rather rotated within the cash universe; (3) bank reserves remained well above the $2.5 trillion ample-reserves threshold the Fed cited in 2024 testimony, declining only modestly through 2025 even as RRP drained. The Fed's December 2023 FEDS Note Money Market Fund Repo and the ON RRP Facility documented this rotation in detail, confirming that the post-2022 RRP drawdown reflected cash-management decisions by money funds rather than a redirection of liquidity into risk assets. The cleanest mechanism implication for SPY is that RRP-versus-SPY is not a useful directional pair, because the marginal dollar drained from RRP did not flow into equities. The mechanism analysis points toward a multi-instrument liquidity composite as the right frame for future policy regime reads. The 2023 SVB and Signature episodes also provided cleaner stress-test evidence: when the regional banking strain hit on March 10, 2023, RRP balances barely moved on the day, while equities sold off and short-end rates whipped. The episode demonstrated that the buffer was not a release-valve mechanism but a stable cash-management facility, and the analytical implication is that flow-of-funds analysis between RRP and risk assets requires accounting for money-fund operational behaviour rather than treating RRP as a fungible liquidity reservoir.

The Convex Net Liquidity Impulse (CNLI) treats RRP as one of five Fed-balance-sheet inputs (Fed total assets, bank reserves, Treasury General Account, RRP, and currency in circulation), and constructs net liquidity as Fed assets minus TGA minus RRP. RRP carries inverted weight inside the composite: rising RRP subtracts from net liquidity, falling RRP adds back. The peak December 30, 2022 RRP at $2.554 trillion subtracted from CNLI in mechanical terms, but CNLI did not flag a stress-regime classification through 2023-2025 because the offsetting flows (reserves stable, TGA in normal range, Fed assets declining at a measured pace) kept net liquidity in the cyclical-neutral range. The April 2026 RRP of $0.7 billion adds essentially zero to net liquidity, leaving CNLI's classification dependent on the other four inputs. The cleanest historical lesson from the RRP-versus-SPY relationship is that CNLI is the better composite for the liquidity-versus-equity question; RRP in isolation is a misleading single-variable indicator, because money-fund cash management decisions during 2023-2026 dominated the simple liquidity-flow interpretation. The composite frame is the only consistent way to read post-2014 Fed-balance-sheet dynamics against equity performance, because each individual leg carries policy-driven distortions that the composite filters out. The CNLI weighting is also dynamic across the rate-policy cycle: during QE phases, Fed total assets carry the largest weight; during QT phases without an RRP buffer (post-2025), bank reserves carry the largest weight; during QT phases with an active RRP buffer (2022-2024), RRP carries a partial weight that reflects its role as the dominant absorption mechanism.

With RRP at zero, the policy regime shifts in three ways. First, any additional QT presses directly on bank reserves rather than on the RRP buffer. The Fed slowed QT in June 2024 (capping monthly Treasury runoff at $25 billion from $60 billion) and ended QT in mid-2025 explicitly to manage the post-buffer transition. Second, money-market rates have less downside protection: with RRP no longer the floor mechanism that absorbed cash at the awarded rate, intraday money-market rate dispersion has widened modestly through 2025-2026, with the SOFR-IORB spread occasionally trading negative (which is the signal that reserves have approached the lower bound of ample). Third, quarter-end repo-market stress is more likely to surface in SOFR spikes rather than RRP collapses: the December 2024 quarter-end saw SOFR trade up to IORB plus 8 bps versus the typical IORB minus 3 bps, the largest dispersion since September 2019. The April 2026 configuration (RRP at zero, reserves at $3.0 trillion, SPY at $568.40) marks the live test of the post-RRP regime, and the watches that matter are SOFR-IORB spread, the Standing Repo Facility take-up, and bank reserve aggregates rather than the RRP balance itself. The September 17, 2019 repo blow-up (SOFR briefly hit 5.25 percent against a fed-funds target of 2.00-2.25 percent) remains the canonical example of what reserve scarcity looks like, and the post-RRP regime is the first time since that episode where reserves are the binding liquidity constraint without a buffer mechanism in place.

RRP-versus-SPY at the multi-year horizon is a falsified directional pair: $2.55 trillion of RRP drained without producing the equity weakness the simple liquidity thesis predicted. The actionable framework is to treat RRP as one input within a broader net-liquidity composite (CNLI), and to read the RRP move alongside the bank-reserves move and the TGA move before drawing any equity-direction conclusion. The most informative episode of RRP-versus-SPY divergence was Q1 2023: the regional banking stress drove a brief SPY drawdown (-7.8 percent peak-to-trough in March) while RRP traded sideways at $2.2 trillion, indicating that the equity stress was credit-led rather than liquidity-led. The 2024 SPY rally during a $1 trillion RRP drawdown indicated the opposite: equity strength independent of the liquidity drain. The pair's signal value is therefore in the disagreement events, not in the steady-state alignment. The horizon for the signal is one to three months, and the data cadence is daily for RRP and continuous for SPY, which makes the pair particularly useful for tactical liquidity-versus-equity calls during quarter-end episodes when RRP and SOFR move together against the broader cyclical pattern. Convex publishes the daily RRP percentile, the bank reserves percentile, the TGA percentile and the CNLI composite reading on the cross-asset dashboard, providing the multi-input frame that any single-series reading needs to be interpreted within rather than read in isolation.