Skip to main content

Reverse Repo (RRP)

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,179 words

The Federal Reserve's Overnight Reverse Repurchase Agreement facility (ON RRP, usually shortened to "RRP" or "reverse repo") is a tool the Fed uses to put a floor under short-term interest rates. In an ON RRP, the Fed sells a Treasury security from its portfolio to an eligible money-market counterparty and agrees to buy it back the next day — effectively borrowing cash overnight from that counterparty at a pre-set rate. For markets-watchers, the RRP matters less as a rate tool and more as a liquidity reservoir: a line on the Fed's balance sheet whose rise and fall reshapes how much cash sloshes through the system, which is why it sits at the center of the popular "net liquidity" framework. Its core tension is that it is genuinely a plumbing instrument with a precise monetary-policy job, yet it has been press-ganged into service as a market-timing signal whose predictive value is far weaker than its devotees claim.

How it works

The Fed's trading desk at the New York Fed runs ON RRP operations every business day at a rate set by the FOMC. Eligible counterparties — primarily money market mutual funds (MMFs), plus government-sponsored enterprises (Fannie/Freddie/FHLBs), primary dealers, and banks — park cash with the Fed overnight and earn the offering rate (NY Fed RRP FAQ; Liberty Street Economics, 2022).

The rate sets a floor because any of these counterparties should be unwilling to lend privately at a rate below what the risk-free Fed offers. MMFs are the key user: they cannot hold reserves and earn IORB (interest on reserve balances) like banks can, so the RRP gives them a Fed-backed alternative when private repo and bill yields fall too low. The offering rate has historically sat a few basis points below IORB, bracketing the effective fed funds rate inside the Fed's target band.

Mechanically, an ON RRP does not change the size of the Fed's balance sheet — it shifts a liability from "reserve balances" to "reverse repo obligations" while the trade is outstanding (NY Fed). Each counterparty may submit one bid up to a per-counterparty cap (raised to $160 billion in 2021). Note the direction confusion: from the Fed's perspective this is a reverse repo (it sells/repurchases collateral); from the counterparty's perspective it is a repo (they lend cash against collateral).

How it's used in practice

Two distinct audiences use the RRP number.

Policy economists read it as a gauge of excess liquidity. A large RRP balance means MMFs have more cash than the private market profitably absorbs, so it pools at the Fed. As of late 2024–2025 the balance fell toward near-zero, which the Fed and analysts treated as a signal that "abundant" reserves were normalizing and that QT was approaching its endpoint (Wolf Street, Nov 2024).

Macro traders fold RRP into a net liquidity proxy, most commonly: Fed total assets (WALCL) − Treasury General Account (TGA) − RRP. The logic is that QT shrinks Fed assets (drains liquidity), but if RRP also drains — i.e., MMFs pull cash out of the Fed and redeploy it into Treasury bills and private repo — that released cash offsets QT. This is precisely what happened in 2023–2024: as the RRP fell from its peak, the cash returning to the system cushioned the impact of QT, and bank reserves stayed broadly stable (roughly $3.3–3.5T through 2023–early 2024) even as the Fed shrank its balance sheet by about $2.2T from its mid-2022 peak (Brookings; ING, 2024). The Fed itself attributes the stability of reserves during runoff largely to the offsetting ~$1.8T decline in the ON RRP.

Adoption, debate & evidence

The RRP went from obscure plumbing to a financial-Twitter staple between 2021 and 2023. Take-up surged from under $1B in early 2021 to just under $2T by end-2021, peaking around $2.55T on a single day (Dec 30, 2022) before draining to a few hundred billion by 2024–25 (Liberty Street; Wolf Street, Aug 2024). Why it grew so large — and whether that was a sign of dysfunction — was itself debated (Bank Policy Institute).

The contested part is the net-liquidity-drives-stocks claim. Proponents cite a very high level correlation (one widely-shared figure put net liquidity's correlation with the S&P 500 near 0.95). Skeptics — including liquidity analyst Michael Howell, who otherwise champions liquidity analysis — warn this is a textbook spurious-correlation trap: both series are non-stationary and trend upward, so level correlations are misleadingly high, and the relationship should be tested in changes, not levels, before being trusted (Capital Wars / Michael Howell). The cleanest counter-evidence: from late 2022 to early 2026, QT removed ~$2.1T of assets while ~$2.4T drained from the RRP, so net liquidity barely moved — yet the S&P rose dramatically over the window (eco3min Net Liquidity dataset). A flat input cannot have caused a large move. The honest verdict: net liquidity is a useful descriptive lens on co-movement and a reasonable check on QT's true bite, but its track record as a predictive market-timing indicator is unproven and likely overstated.

Strengths & limitations

Strengths. As a plumbing read it is excellent and low-controversy: a high RRP genuinely signals excess cash and a soft floor regime; a near-zero RRP genuinely signals that the system's liquidity buffer is thinner and that money-market rates (SOFR, TGCR) deserve closer watching. It is daily, public, and clean (FRED series RRPONTSYD).

Limitations. (1) The causal chain from RRP/net liquidity to equity prices is unestablished; the eye-catching correlations are largely spurious. (2) The RRP can no longer "save" markets the way it did — once it is drained near zero, the cushion it provided to QT is gone, so the same indicator that looked bullish on the way down is exhausted at the bottom. (3) It is one leg of a three-part identity (assets, TGA, RRP); TGA swings around debt-ceiling episodes and tax dates can dominate, so reading RRP alone is misleading. The #1 misuse is treating the level-correlation with the S&P as a tradable signal and front-running it without testing in changes — a classic over-fit to the 2021–2023 regime that broke afterward.

Sources