Jeff Huther
DataBank
The Federal Reserve recently released data of discount window and Bank Term Funding Program financing that shows how the facilities were used following the failure of Silicon Valley Bank in March 2023. While aggregate data on borrowing from these two facilities has been available since 2023, the new transaction-level data provides a more complete picture of who borrowed, how much they borrowed and how long the loans were outstanding. In this DataBank we summarize the findings of our analysis. We show for example, eligible institutions did not rush to borrow from the BTFP, and while a lot of public attention has been paid to bank usage, we find that credit unions were also active borrowers from the Fed.
The data
The Dodd-Frank Act mandated the release of transaction-level data for Fed lending programs within two years of the transaction. The Fed chose to release the full data set for the BTFP when the program formally ended in March 2025 (9,812 transactions) and its quarterly releases of DW lending now includes 2023Q1 (3,525 transactions).
The discount window and BTFP background
The discount window is a long-standing Fed facility that helps the Fed perform its lender of last resort function. In the wake of the failure of SVB, the Fed created the BTFP which had three notable differences from the DW: It accepted only high-quality collateral at par rather than discounted market value, it allowed the borrower to set the maturity (up to a year and the ability to prepay at no cost) and it used a market-based interest rate. Over the year of BTFP lending, these differences influenced the pattern of bank usage of the two facilities.
Figure 1: BTFP outstanding and the interest on reserves – BTFP rate spread
The one-year swap rate began to fall in late 2023 when market participants brought forward their expectations of Fed interest rate cuts. As shown in Figure 1, the positive rate spread led to a large increase in BTFP usage that came to an end when the Fed changed the rate formula in January 2024. Lending ceased one year after SVB failed and the last loans were repaid on March 6, 2025, a few days before the program officially ended.
Key findings
- Differential timing of DW and BTFP borrowing: Three banks made use of the BTFP the same day it was opened. The level of interest from banks and CUs peaked in the second week of availability, which is not surprising given that the BTFP was a new program, with a novel structure. As shown in Figures 2A and 2B, the number of DW borrowers peaked on the first business day after the SVB failure and then gradually declined while BTFP usage spiked on the Friday two weeks after the failure, suggesting precautionary interest ahead of the weekend and time needed to gain familiarity with the program. A direct comparison of borrowing amounts is complicated by loan terms — most BTFP loans were term loans while most DW loans were overnight.
Figure 2A: Number of BTFP borrowers
Figure 2B: Number of DW borrowers
- Large-scale DW borrowing from the Fed in March 2023 was out of need but most transactions were relatively small: While the actual motivation for borrowing from the Fed is not specified in the data, the long DW data series show that many banks regularly test their DW access via small loans (e.g., <$10,000) for short terms (e.g., one day). As a result, the loans to the failed banks (First Republic Bank, Signature Bank and SVB) and their successors represented over 80% of all DW borrowing by volume but only 3% of all transactions between March 13, 2023, and March 31, 2023. Figure 3 shows the amount each of these banks borrowed from the DW in the first quarter of 2023.
Figure 3: DW Borrowing of failed and failing banks in Q1 of 2023
- Most BTFP borrowers prepaid their loans: Most banks and CUs took out BTFP loans for the fully available maturity (one year) but many prepaid them. Borrowers pre-paid 80% of their BTFP loans, usually when the borrowing rate exceeded IOR. CUs were slightly more aggressive in taking advantage of the prepayment option. They asked for slightly longer loans (than banks), but their actual loan lengths were shorter, as shown in Figure 4.
Figure 4: BTFP loan maturities
- No indication that BTFP affected DW collateral selection: While the DW takes almost any asset as collateral, BTFP collateral was limited to Treasury and Agency securities. Therefore, the composition of collateral pledged to the DW could have shifted towards riskier assets after the creation of the BTFP. The reasoning is that banks and CUs would have moved their high quality, lower risk collateral to the BTFP, and replaced them with higher risk assets at the DW. As shown in Figure 5, this shift did not occur and the share of BTFP-eligible collateral that was posted to the DW was not noticeably lower in the latter half of March 2023 compared to the preceding period.
Figure 5: Share of BTFP-Eligible collateral posted to the DW
Conclusion
This DataBank has covered just a few key highlights and conclusions to be drawn from the loan level data for the BTFP and DW from March 2023. While the focus of usage of Fed facilities is usually on banks, CUs were also borrowers at these two facilities. As the Fed releases additional DW data, it will be possible to construct a more complete story about how the BTFP complemented the Fed’s traditional lending facility to address emerging liquidity risks in the funding markets.
Jeff Huther recently retired as VP for banking and economic research in ABA’s Office of the Chief Economist. A former Federal Reserve economist, Huther will continue to contribute occasional research and analysis to ABA.
For additional research and analysis from ABA’s Office of the Chief Economist, please see the OCE website.
