Bank reserves are the deposits commercial banks hold in their accounts at the Federal Reserve — the ultimate settlement asset of the U.S. financial system and the largest single liability on the Fed's balance sheet. "Net liquidity" is a popular macro-trader construct that tries to isolate the portion of the Fed's balance sheet actually circulating in the financial system by subtracting two large, non-reserve liabilities — the Treasury General Account (TGA) and the overnight reverse repo facility (ON RRP) — from the Fed's total assets. The core tension: reserves are a precise, well-defined plumbing quantity with real consequences for funding markets, whereas "net liquidity" is a proxy that has been correlated with risk-asset prices in some windows but is contested, regime-dependent, and frequently mis-stated.

How it's calculated / formed

Reserves are created mechanically by the Fed's own operations. When the Fed buys a security (QE) or lends to a bank, it credits that institution's reserve account; when it sells securities or lets them roll off (QT), reserves fall. Critically, reserves are a residual: total Fed assets must equal total Fed liabilities, so for a given asset level, reserves fall when other liabilities rise. The two big swing factors are:

• TGA (FRED WTREGEN) — the Treasury's checking account at the Fed. When Treasury issues debt or collects taxes faster than it spends, cash moves from bank reserves into the TGA, draining reserves. Spending it back down adds reserves.
• ON RRP (FRED RRPONTSYD) — mainly money-market funds parking cash at the Fed overnight. Cash in RRP is sterilised out of the banking system; when RRP drains (2023–2024), that cash flows back into reserves.

The widely used net liquidity formula (popularised by Arthur Hayes and in the same spirit as 42 Macro / Darius Dale work) is:

Net Liquidity = WALCL − WTREGEN − (RRPONTSYD × 1,000)

where WALCL is total Fed assets. The RRP series is scaled because FRED reports it in different units. All three series update weekly from FRED. Note this is an approximation of system liquidity, not an official Fed statistic — actual reserve balances are reported directly as "reserve balances with Federal Reserve Banks" on the H.4.1 release.

How it's used in practice

Two distinct uses, often conflated:

1. Funding-market / plumbing read (the rigorous use). Reserve scarcity is monitored to anticipate stress in repo and the fed funds market. As reserves drop toward the system's "lowest comfortable level," money-market rates (SOFR, the SOFR–IORB spread, fed funds) become increasingly sensitive to TGA swings, settlement dates, and Treasury issuance. Sharp upward spikes signal that QT or a TGA rebuild has pushed reserves below comfort — historically a precursor to Fed intervention (standing repo facility, ending QT, restarting purchases).

2. Risk-asset directional overlay (the popular, contested use). Discretionary macro and crypto traders chart net liquidity against the S&P 500 / Bitcoin as a directional tailwind/headwind: rising net liquidity = supportive, falling = a drag, with a lag commonly cited at roughly two weeks. It is used as a regime/context input — a reason to lean long or de-risk — not a precise timing trigger.

Adoption, debate & evidence

The plumbing interpretation is mainstream and well-evidenced. The September 2019 repo spike is the canonical case: aggregate reserves had fallen to a multiyear low (under ~$1.4 trillion), and a confluence of corporate tax payments and large Treasury settlements pushed overnight repo from ~2% to as high as ~10% intraday (SOFR jumped from 2.43% on Sep 16 to 5.25% on Sep 17), forcing the Fed back into open-market operations. This demonstrated that reserves can become scarce and unevenly distributed even under an "ample reserves" framework (New York Fed; OFR; Yale Journal of Financial Crises).

The net-liquidity-predicts-stocks claim is genuinely contested and frequently overstated. Widely circulated figures of "0.85–0.95 correlation with the S&P 500" should be treated skeptically:

• Both net liquidity and the S&P 500 are non-stationary, upward-trending series, so a high level correlation is partly spurious — two things trending up will correlate regardless of causation.
• The correlation is highly regime-dependent. One independent analysis (eco3min, using FRED data) found net-liquidity↔S&P level correlation of ~0.81 in 2020–2022 but roughly zero (−0.03) in 2023–2024, even as the index rose ~40%. The famous "0.91" stems from a specific mid-2021 rolling window and does not hold out-of-sample.
• A subtle transmission point: during 2022–2025, QT drained mostly RRP, not reserves, so balance-sheet shrinkage hit system reserves far less than 1:1. Once RRP is exhausted, further QT falls directly on reserves — meaning the same headline net-liquidity number can imply very different stress depending on which liability is moving.

Bottom line: as funding-stress radar, reserves are first-rate. As a mechanical predictor of equity direction, net liquidity is a context overlay with weak, unstable standalone edge — no peer-reviewed study establishes it as a reliable timing signal.

Strengths & limitations

Strengths. Reserve tracking gives genuine early warning of repo/funding stress and pending Fed policy pivots — a real, mechanical linkage. The component decomposition (TGA, RRP) explains why reserves move, which is more informative than the headline alone.

Limitations. (1) The S&P correlation is unstable and partly spurious; treating a level chart overlay as causal is the single most common misuse. (2) The simple WALCL−TGA−RRP formula ignores other liabilities and the reserves-vs-RRP composition that determines actual stress. (3) Regime dependence: works best when liquidity is the dominant market driver (e.g., 2020–2021); fails when earnings, rates, or AI/secular themes dominate (2023–2024). (4) Lags and data revisions make precise timing unreliable.

Sources

• Federal Reserve, Federal Reserve Liabilities (reserve balances, TGA, RRP, currency) — federalreserve.gov/monetarypolicy/bst_frliabilities.htm
• FRED series: WALCL (total assets), WTREGEN (TGA), RRPONTSYD (ON RRP) — fred.stlouisfed.org
• New York Fed, Reverse Repo Operations and The Market Events of Mid-September 2019 (Afonso et al.)
• Office of Financial Research, Anatomy of the Repo Rate Spikes in September 2019 (OFR WP 23-04); Yale Journal of Financial Crises, "Anatomy of the Repo Rate Spikes"
• eco3min, Net Liquidity Index (2003–2026) — period-by-period correlation analysis; flags spurious/regime-dependent correlation [contested-claim source, used for the skeptical case]
• Richmond Fed, The Borrower of Last Resort: What Explains the Rise of ON RRP Facility Usage? (2021)
• Arthur Hayes / 42 Macro (Darius Dale) — popularisers of the net-liquidity formula [practitioner, not academic; cited for attribution only]

> Dispute flagged: the "0.85–0.95 net liquidity ↔ S&P 500 correlation" widely cited by practitioners (e.g. GuruFocus, crypto commentators) is not corroborated out-of-sample and is partly a spurious-trend artifact. The plumbing/reserve-scarcity claims are well-sourced; the equity-prediction claim is weak and regime-dependent.