Inside the BTFP Subsidy
Banks and credit unions earned free money by borrowing from the Fed's "emergency" program and then lending it back to the Fed (or other banks) at higher rates.
It has been two years since the banking “crisis” that claimed Silvergate, Signature, Silicon Valley Bank, and First Republic—and one year since the Federal Reserve stopped extending new “emergency” loans under the Bank Term Funding Program (BTFP). The final loans were repaid on March 11, 2025, and the Fed has since published a spreadsheet with the details. While the BTFP was not designed to print money for banks and credit unions, it effectively did so in November 2023-January 2024 before Fed officials changed the terms of the program.
The BTFP allowed banks to borrow up to 100% of the face value of bonds issued or guaranteed by the U.S. Treasury. In practice, this mostly meant mortgage-backed securities (MBS) and collateralized mortgage obligations (CMOs) from Fannie Mae, Freddie Mac, and Ginnie Mae. Until January 25, 2024, the interest rate on BTFP loans was set daily at the 1-year overnight index swap (OIS) rate, plus 0.1 percentage point. There was no prepayment penalty. The goal was to allow lenders with unrealized losses on their bond portfolios to be able to get cash at relatively attractive terms without having to sell assets that would eventually be money good.
This may have been helpful during the spring of 2023, as depository institutions borrowed almost $100 billion between mid-March and early June. After that, however, things had stabilized, and there was less of a point to the program. Yet by November, something strange had happened: the terms of the BTFP had created an opportunity for anyone willing to exploit it by borrowing cash from the Fed via the BTFP and lending it back to the Fed by holding reserves, or to other banks by lending intraday or overnight. Lenders did this in size, not only refinancing their existing obligations, but increasing their net BTFP borrowing to a peak of almost $170 billion. Intriguingly, tax-exempt credit unions were among the most aggressive users of the facility for this purpose.
What follows is a closer look into how this happened, and which lenders did it.
The Initial Incentive to Borrow from the BTFP
Banks, credit unions, and other depository institutions finance most of their assets by borrowing from depositors, who think of their short-term loans as equivalent to money. While most of those deposits can be redeemed for cash—or wired to another financial institution—more or less on demand, most of the corresponding assets cannot. With the exception of reserves held by depositories at the Fed, selling or borrowing against assets under duress is expensive. Buyers will demand steep discounts and lenders will demand haircuts and charge usurious interest rates. Since it is impossible to plug a deposit hole by relying on “good” assets alone, even a sound bank would collapse if enough depositors ever decided to cash out at the same time, for whatever reason.
Banks like this setup because—most of the time—they can profit by paying less for deposits than they earn from their assets. And if things ever get bad enough, the rest of society will come to their rescue.1 This is a strange way to construct a financial system, and I have repeatedly argued that there are alternatives. Nevertheless, it is the system we have, which is why the Fed and other central banks try to follow the classic advice on managing financial crises: lend freely against good collateral at a penalty rate.
The BTFP was supposed to follow this model. In many ways, it was far more generous than the Fed’s default emergency lending program (Primary Credit offered via the discount window) as explained in this presentation from the time, but the flipside was that the BTFP only lent against assets with zero default risk.
That structure made sense given what was happening in 2023H1. Banks had bought $2 trillion worth of U.S. Treasurys and mortgage bonds in 2020-2021. Those were safe from credit risk, which made them attractive given the regulatory environment and the cost of deposits at the time. But as the Fed began raising short-term interest rates in 2022, the bond portfolio ended up yielding far less than what was available in the money markets. (Equivalently, the market value of the bonds fell, because the fixed schedule of coupon payments had become less attractive relative to what could be earned by lending overnight to the Fed.)
That would not have been a problem if depositors had kept their cash in low-yielding bank accounts. Unfortunately for the banks, however, larger and more sophisticated depositors almost immediately began pulling their money out of the banking system and shifting it to money market funds that held Treasury bills and/or lent overnight to the Fed. Deposits above the Federal Deposit Insurance Corporation’s (FDIC) insurance threshold of $250,000 fell by $820 billion between 2021Q4 and 2022Q4. That was before the bank failures that started months later.
Banks evidently decided that paying up for deposits was less appealing than selectively replacing the lost financing in the market at much higher rates. Bank borrowing from the Federal Home Loan Banks (FHLBs) jumped by $400 billion between 2021Q4 and 2022Q4, which, along with the rise in insured deposits, helped offset the flight of uninsured deposits. Even this might have seemed relatively benign: FHLB lending had plunged during the pandemic, so the rebound through the end of 2022 could be viewed as a return to normalcy.
By 2023, however, relatively-large banks started getting into serious trouble and borrowing from the FHLBs surged even more. That is when the Fed stepped in, initially via a spike in discount window lending in early March and then with the BTFP on March 13. These alternative financing sources are included in what the FDIC calls “other borrowed money”. Borrowing from the FHLBs normalized by 2023Q2, with Fed lending helping to fill the gap.
So what changed?