In September 2019, the plumbing of the U.S. financial system seized in a way that surprised almost everyone watching it. The rate on overnight repurchase agreements, or repos, the market where banks and funds borrow cash overnight against Treasury collateral, spiked from around 2% to nearly 10% in a matter of hours. Cash that was supposed to be abundant simply stopped moving. The Federal Reserve had to inject tens of billions of dollars through emergency operations to force the rate back down.
What made the episode strange was not that cash got tight. Money-market plumbing gets tight around quarter-end and tax dates fairly often. What made it strange was that a facility designed to prevent exactly this kind of spike was, at that moment, not doing its job. The discount window, the Fed's oldest lending tool, sat open and available the entire time. Banks with plenty of Treasury collateral could have borrowed from it at a modest penalty and lent that cash into the repo market at 10%, pocketing an enormous risk-free spread. Almost none of them did.
That refusal is the whole story. Understanding why banks would rather watch a fire burn than reach for the fire extinguisher the Fed hands them is the key to understanding both the discount window and its newer cousin, the standing repo facility.
What these two facilities actually are
A backstop facility is a standing offer from the central bank to convert safe collateral into cash on demand, at a price. The bank pledges assets it already owns, most often Treasury securities, and receives cash in return, agreeing to unwind the trade shortly after. The price is a penalty rate set slightly above where healthy markets clear, so that in normal times nobody uses the facility, and in stressed times it caps how high short-term rates can go.
The discount window does this against a broad range of collateral, including loans and less liquid assets, and lends for periods up to 90 days. It has existed since 1913. The standing repo facility, or SRF, launched in July 2021, does something narrower and faster: it swaps cash for Treasury and agency collateral overnight, settling same-day, with counterparties that include primary dealers and eligible banks.
The design logic is identical. If a bank can always get cash at, say, the top of the Fed's target range, then no rational lender should ever be able to charge much more than that in the open market. The facility rate becomes a ceiling. The repo rate cannot spike to 10% if any participant can borrow from the Fed at 5% and lend the proceeds out. Arbitrage should crush the gap almost instantly.
The word "should" is carrying an enormous amount of weight in that sentence, and 2019 showed us why.
The mechanism that breaks the ceiling
The ceiling only holds if banks will actually use the facility. Here is where intuition fails: economists model these tools as a price cap, but the tools are not a price. They are a decision made by a risk officer who has to weigh the arbitrage profit against a cost that does not appear on any rate sheet. That cost is stigma, the market's assumption that a firm which borrows from the central bank must be in trouble.
The causal chain runs like this. A bank taps the discount window. Borrowings show up, with a lag, in aggregate Fed data and, historically, sometimes leaked through market chatter or counterparty behavior. Other banks, seeing evidence that a peer needed emergency cash, pull their own funding lines from that bank as a precaution. The act of reaching for the backstop becomes the trigger for the run the backstop was meant to prevent.
So the risk officer facing the September 2019 spike did the math differently than the economist did. Borrowing at the window to earn a few days of spread was worth perhaps a few million dollars. The possibility that the borrowing would be noticed, and would mark the firm as weak, carried a tail risk measured in the firm's survival. No spread compensates for that. So the bank sold assets instead, or simply sat on its cash, and the rate spiked because the arbitrage that was supposed to be automatic never happened.
The practical takeaway is that a liquidity backstop is only as strong as the willingness to use it, and willingness collapses at exactly the moment liquidity is scarcest. A fire extinguisher bolted to the wall with a sign reading "using this means everyone will assume your house was already burning" is not a fire extinguisher. It is decoration.
Why stigma is a coordination failure, not a character flaw
It is tempting to blame individual banks for cowardice, but the stigma problem is structural. It is a coordination failure: a situation where the collectively rational action differs from the individually rational one, and no single participant can afford to move first.
If every bank borrowed from the window during stress, no single borrower would stand out, and stigma would evaporate. Borrowing would signal nothing because everyone did it. But no bank wants to be the first, or the only, name on that list. So each waits for others to go first, and none do. The equilibrium is universal abstention even when universal participation would be safe.
This is the same structure that produces bank runs among depositors. A depositor who trusts a bank still withdraws money if she believes others will withdraw, because the last person in line gets nothing. Being right about the bank's health does not protect you from the consequences of everyone else's fear. Facility stigma is that logic applied one layer up, among the banks themselves.
The 2023 regional bank failures made this vivid. Silicon Valley Bank held a large book of Treasury and agency securities that were eligible collateral. In principle it could have pledged them for cash and met deposit outflows without selling anything at a loss. In practice it sold roughly $21 billion of securities at a realized loss, which announced its distress to the market and accelerated the run. Reaching for the backstop felt more dangerous than the fire sale, so the fire sale is what happened.
For a reader watching bank stocks or holding deposits, the lesson is that a bank's stock of eligible collateral tells you what it could do, not what it will do. The gap between the two is stigma, and it widens under stress.
What the SRF was built to fix, and what it cannot
The standing repo facility was designed with the stigma problem explicitly in mind. Because it operates against only the safest collateral, Treasuries and agencies, and settles as an ordinary market transaction rather than an emergency loan, the Fed hoped it would carry less of the discount window's taint. Using the SRF is meant to feel like a trade, not a confession.
Whether that framing survives real stress is the open question. Stigma is a belief held by other market participants, and the Fed cannot legislate what conclusions rival banks draw when they see a firm using a facility. A tool can be redesigned; the market's instinct to read backstop usage as a distress signal is harder to redesign. 
There is also a mechanical limit. The SRF only accepts high-quality collateral, so it stabilizes the repo rate and the market for Treasuries. It does nothing directly for a bank whose problem is that it holds assets nobody wants at any price. In a genuine credit panic, the fire sale moves to whatever the facility will not take, and the ceiling holds only over the narrow slice of collateral the Fed accepts.
The practical implication for anyone reading the plumbing is to watch two numbers together. When short-term funding rates climb toward the facility ceiling, ask whether usage of the facility is actually rising to meet it. If rates press against the ceiling while usage stays near zero, stigma is winning, and the backstop is a ceiling in name only. That divergence, not the headline rate, is the real stress gauge.
The one thing to remember
A central bank backstop stops a fire sale only if banks will use it before they sell, and stigma is the force that reliably delays that decision until the selling has already begun. The facilities are circuit breakers wired to a switch that market participants are afraid to flip. That is why the Fed keeps redesigning them, why the SRF exists at all, and why no design has yet fully solved a problem that lives in beliefs rather than in rates.
