The question I got most during the 2023 grind higher was why bitcoin kept working while the Fed was running quantitative tightening. On paper the central bank was letting tens of billions of dollars of bonds roll off its balance sheet every month, which is supposed to drain dollars from the system, and risk assets are supposed to struggle when that happens. The honest answer was that the dollars were coming from somewhere else, and the somewhere else was a facility most crypto people had never looked at. The Fed's overnight reverse repo facility, ON RRP in the official language, spent about two years releasing cash back into markets at a pace that swamped what QT was removing, one quiet daily print at a time.
What the facility actually does
The mechanics are simpler than the name suggests. Every business day, money market funds and a short list of other eligible counterparties can lend cash to the Federal Reserve overnight. The Fed posts Treasury securities as collateral and pays an overnight rate that the FOMC sets alongside its main policy rate. The next morning the trade unwinds and the fund can roll it again. As overnight parking spots go it is about as safe as anything on earth, since the borrower is the institution that issues the currency.
Why would trillions sit there? Because after the 2020 and 2021 wave of QE and fiscal stimulus there was far more cash in money market funds than there were Treasury bills to absorb it. Bill supply was constrained for long stretches, partly because debt ceiling standoffs kept forcing the Treasury to cut issuance. So the funds took the Fed's standing offer. Usage went from near zero in early 2021 to roughly two and a half trillion dollars at the peak in late 2022, a number that still looks like a typo when I write it out.
The property that matters is that RRP cash is inert. While it sits at the Fed it funds no loans, buys no bills, and backs no positions. It has effectively stepped outside the financial system for the night, every night. So the facility behaves like a reservoir. Cash flowing in pulls liquidity out of markets, and cash flowing out releases it back.
Why the drawdown worked like easing nobody announced
There is one piece of accounting worth memorizing here. Bank reserves, the deposits commercial banks hold at the Fed, roughly equal the Fed's total assets minus the Treasury's checking account, the TGA, minus the RRP balance. Currency in circulation and a few smaller items sit in the equation too, but they move slowly, so the three fast movers are the balance sheet, the TGA, and the RRP. Reserves are the raw material of market liquidity. When banks are flush with them, balance sheet is cheap, funding is easy, and risk appetite tends to follow with a lag.
QT shrinks the asset side, which on its own shrinks reserves. But if the RRP is draining at the same time, the two effects push against each other, and through most of 2023 the drain was the bigger force. Money funds pulled cash out of the facility to buy the flood of new T-bills the Treasury issued once the debt ceiling fight resolved, mostly because bills yielded a little more than the Fed was paying. The Treasury spent what it raised, the money landed in bank accounts across the economy, and reserves rose. The Fed removed hundreds of billions of dollars through QT that year and banking system reserves still finished the year higher than they started. The RRP drawdown is more or less the entire explanation for that oddity.
Crypto felt the effect before most people could name the mechanism. Bitcoin trades like a high-beta claim on dollar liquidity, whatever story we tell ourselves about adoption. When reserves rise, stablecoin supply tends to expand, market makers run bigger books, perp funding loosens, and the marginal buyer shows up more often. I would not defend a tick-for-tick correlation, since ETF flows, exchange drama, and a hundred crypto-native things dominate any given week. Over a horizon of months, though, the direction of net liquidity has been one of the more honest macro inputs I track, and during the QT years the RRP was the swing variable inside it.
Pulling the number and reading it properly
This is one of the easiest macro series to follow because the Fed publishes it every day. My routine looks like this.
- RRP usage: the New York Fed posts each day's operation result on its temporary open market operations page in the early afternoon New York time. FRED mirrors it as the daily series RRPONTSYD.
- TGA: the Daily Treasury Statement carries the cash balance with a one day lag, or use the weekly WTREGEN series on FRED if daily is more than you need.
- Reserves: the Fed's H.4.1 release comes out weekly, and the WRESBAL series on FRED carries the level.
- Net liquidity sketch: take WALCL, the Fed's total assets, subtract the TGA and the RRP, and watch the direction over four to eight weeks rather than any single print.
The failure modes matter more than the sources, so here are the three I see most. The big one is reading RRP in isolation. A falling RRP releases cash, but if the TGA is being rebuilt at the same time, say after a debt ceiling deal or around large tax dates, the Treasury is soaking that liquidity right back up and the net effect can be flat or negative. I have watched people call a liquidity bottom off the RRP chart alone while the TGA quietly absorbed everything the facility gave back.
Second, ignore quarter ends when you read the chart. Dealers shrink their repo books over reporting dates, money funds shift cash to the Fed for a night or two, and the RRP prints a spike that reverses within days. That spike is a story about bank regulation and reporting windows rather than a change in the liquidity trend.
Third, the signal changes character when the facility gets close to empty. A fat RRP acts as a buffer that lets QT run without draining reserves. Once the buffer is spent, runoff bites reserves directly, and the thing to watch shifts to repo rates and the reserve level itself. The 2019 repo spike, when overnight rates briefly blew out because reserves got scarcer than the Fed expected, is the historical case study for that regime.
None of this needs a terminal or a data subscription. Fifteen minutes on FRED once a week covers it, and the discipline of reading the three numbers together is worth more than any single source. We fold a version of this net liquidity read into the macro context on Blockcircle for that reason, but the raw series are free and public. If you trade crypto on any horizon longer than a few days, knowing whether the reservoir is filling or draining is cheap information, and it usually tells you more than the press conference does.