What bank reserves are, why there is no fixed dollar threshold for "enough", which indicators actually tell you scarcity is arriving, and why DLI keeps the reserve level out of the score while showing it on the page.
Bank reserves are deposits that commercial banks hold at the Federal Reserve. They are not vault cash, they are not loans, and — importantly — they are not something banks can collectively spend down. An individual bank can move reserves to another bank; the system as a whole can only change its total reserve balance through the Fed's own actions or through the Treasury's account.
That last point is why reserves sit at the centre of any liquidity framework. When the Fed buys a security, it credits reserves into existence. When a security rolls off the balance sheet without reinvestment, reserves are extinguished. When the Treasury collects taxes into the General Account, reserves leave the banking system; when it spends, they come back. The reserve balance is the settlement medium the entire dollar system clears through, which is why its scarcity shows up in overnight funding markets before it shows up anywhere else.
The single most common mistake in reserve analysis is looking for a dollar number below which reserves become scarce. There isn't one, and the history makes that plain. In September 2019 the system hit acute scarcity with roughly 1.4 trillion dollars of reserves. In 2026 the system runs comfortably with a multiple of that. Nobody moved the goalposts — the banking system that the reserves have to support grew, regulatory liquidity requirements changed, and intraday payment volumes rose.
Scarcity is therefore relative by construction. The useful measure is reserves plus the overnight reverse repo balance as a share of total commercial-bank assets: the cash cushion relative to the balance sheet it has to clear. That ratio is what the site tracks as the reserve buffer. Read against it, September 2019 was an 8-to-9-percent event, 14 percent is roughly where a cushion stops being comfortable, and the current reading sits near 12 percent — thin, but with the plumbing calm.
Reserves themselves are a weekly, slow-moving series. The market's opinion about whether they are sufficient updates every morning, in the overnight funding market, and that is where to look.
The first tell is SOFR against IORB. Interest on reserve balances is what the Fed pays banks to hold reserves; it should function as a soft floor for secured overnight lending. When SOFR trades persistently above IORB, someone is willing to pay more for cash overnight than the Fed pays — the definition of reserves being scarce at the margin. The second tell is the tail rather than the median: the 99th-percentile SOFR print blows out at quarter-ends well before the median moves, which makes it an earlier and noisier warning. The third is Standing Repo Facility usage. The SRF exists so that a bank short of cash has somewhere to go that is not the open market; any material usage means someone reached for the backstop.
The pattern in 2019 is worth memorising because it is what these indicators are for. The buffer sat at 8 to 9 percent and drained for months while overnight rates stayed calm — no stress to report, right up until the morning of September 17, when repo printed 10 percent. Scarcity is a state that accumulates quietly and then discharges suddenly. That is why a liquidity framework needs both a stock reading and a flow reading, and cannot get by on either alone.
Reserves enter the DLI score in two forms and are excluded in a third, and the distinction is the single most-argued point about this index.
They enter as flow: the net-liquidity spine is the Fed balance sheet minus the Treasury General Account minus the overnight reverse repo facility, and its change is the reserve impulse — reserves being created or destroyed. They enter as buffer: the reserve-relative gain scales up a draining flow when the cushion is thin, and a bounded scarcity term prices a genuinely thin cushion directly.
What stays out is the raw level. A reserve-level term in the score fails empirically, not aesthetically: reserves-to-GDP correlates with the Chicago Fed NFCI at about −0.745, meaning that across the ample-reserve era the level runs against financial conditions. Put it in the score and the index prints tight precisely when reserves are being rebuilt and conditions are easing. The same argument applies to the Treasury General Account, where the intuition "TGA is high, so liquidity must be tight" is a known-wrong reading — what matters is whether the account is filling or draining.
So the level is shown, not scored. Bank reserves appear on the site as a display-only context indicator, in raw trillions, alongside the buffer ratio. That is the honest arrangement: the number is genuinely informative for understanding where the system is, and genuinely misleading if you let it vote on the score.
There is no fixed number. September 2019 was an acute scarcity event at roughly 1.4 trillion dollars of reserves, while 2026 runs comfortably on a multiple of that, because the banking system, regulatory liquidity requirements and payment volumes all grew. Use the relative measure instead: reserves plus ON RRP as a share of commercial-bank assets, where 14 percent is roughly where comfort ends and 8 percent is the 2019 crisis floor.
SOFR against IORB. Interest on reserve balances should act as a soft floor for secured overnight lending, so a persistent positive spread means someone is paying more for overnight cash than the Fed pays — scarcity at the margin. The 99th-percentile SOFR tail moves earlier still at quarter-ends, and Standing Repo Facility usage confirms that someone reached for the backstop.
Because it anti-tracks conditions. Reserves-to-GDP correlates with the Chicago Fed NFCI at about −0.745, so a level term makes the index print tight exactly when reserves are being rebuilt and conditions are easing. Reserves do enter the score as flow (the net-liquidity impulse) and as the buffer that scales a drain. The raw level stays on the site as display-only context.
No — that reading is a known error. What moves liquidity is whether the account is filling or draining, not how full it is. A high but flat TGA is neutral; a rapidly filling TGA drains reserves regardless of its starting level.
Education
Most retail investors hold a few cached beliefs about Fed policy that don't survive contact with the data. Here are four of the biggest, with what the historical record actually shows for each.
Education
Everything you need to know about the Treasury General Account — how it works, why TGA drawdowns inject liquidity, how rebuilds drain it, historical swings during debt ceiling crises, and how to use TGA data for investment decisions.
Education
A step-by-step guide to reading the Federal Reserve balance sheet (WALCL) — what its components are, what expansion and contraction mean for markets, and how to use it as an investment signal.