Bank Term Funding Program: Experience and Lessons Learned

The Federal Reserve recently released loan-level data on the Bank Term Funding Program (BTFP). This note combines the Fed's disclosures with bank call report data to identify characteristics of banks that took advantage of the BTFP. Furthermore, while the Fed's disclosures did not include information on the market values of the securities that collateralized the Fed's loans, this note estimates the magnitude of the undercollateralization. This estimation is crucial for assessing the extent to which the BTFP transferred risk from borrowing banks to the Treasury, and therefore to taxpayers, through the credit support provided by the Treasury to the Fed for the BTFP.

This analysis indicates that banks that took advantage of the BTFP were predominantly those that had suffered unusually large deposit runoffs following the bank failures in early 2023. Even controlling for the magnitude of deposit runoffs, banks that had suffered more significant unrealized losses on their securities portfolios were more likely to borrow, as were banks for which a larger share of their deposits were uninsured, perhaps because those banks were more apprehensive about potential future deposit runoffs. Furthermore, banks that had borrowed at the discount window in 2022 were more likely to use the BTFP, perhaps because those banks were less deterred by the stigma often attached to borrowing from the Fed.

Because of a flaw in how the program was designed, in November 2023 the loan interest rate fell below the rate that depository institutions could earn simply leaving the loan proceeds on deposit at a Federal Reserve Bank. The resulting arbitrage opportunity induced banks to increase borrowing significantly at the end of the year, even though strains on the banking system had eased. In January, the Fed took action to remove the arbitrage opportunity, and borrowing promptly fell back.

Our estimates of the extent to which the loans made under the BTFP were undercollateralized indicate that in aggregate the shortfall reached more than $20 billion. Because of credit protection provided to the Fed by the Treasury, the Fed probably was not exposed to losses, but the Treasury was, possibly contrary to the legal authority under which the Fed created the BTFP.^[1]

We also estimate that by taking over an undercollateralized loan that had been made to First Republic Bank upon that bank's failure, the FDIC shifted approximately $3 billion in costs that should have been borne by the Treasury onto the banking system.

In sum, the BTFP appears to have operated largely as intended, but at considerable risk and cost. Going forward, the Fed should ensure that banks are prepared and willing to use the regular discount window when the banking system is under strain, which would reduce or eliminate the need for a facility like the BTFP.

Background

The Bank Term Funding Program was opened on Monday, March 13, 2023, to help limit contagion in the banking sector in the wake of the failures of Silicon Valley Bank and Signature Bank. The BTFP was created under the Fed's Section 13(3) emergency lending authority, which the Fed had used repeatedly during the Global Financial Crisis and the COVID-19 pandemic.

On March 11, 2025, the last loan extended by the Federal Reserve under the Bank Term Funding Program was repaid. The next day, one year after the closing of the program, the Federal Reserve Board, as required by law, released the loan-level data for the BTFP (link).

The BTFP provided loans to financially sound depository institutions with maturities of up to one year at the one-year OIS swap rate plus 10 basis points. The loans were collateralized by Treasuries, agency debt and agency MBS. At its peak, the BTFP was providing over $165 billion in loans. The program made nearly 9,000 loans to about 1,327 borrowers. Because the loans could be repaid without penalty, many of the loans were to repeat customers that were repaying early and borrowing again at a lower rate, in some cases for several days in a row.

What made the program extraordinary was that the Fed lent up to the full par value of the securities pledged as collateral at a time when the fair values of the securities held by the banks were on average about 20 percent underwater. Under its regular lending programs, the Fed assigns a lendable amount to each item of collateral equal to the fair value of the collateral minus a haircut. Because the BTFP collateral had par values above market value, the loans were unsecured to the extent of the difference between the par and market value. The Treasury provided the Fed $25 billion in credit protection using the Exchange Stabilization Fund, presumably to ensure that the Fed was not exposed to losses on the unsecured parts of the loans.

Even though the borrowers were depository institutions and therefore had access to liquidity at the discount window, the loans were provided under Section 13(3) of the Federal Reserve Act, the Fed's emergency authority to extend loans to a broad-based group of institutions or individuals. In the past, 13(3) has almost always been used to authorize lending to nonbanks, not banks. The Fed may have used 13(3) because the term of the loans was one year, beyond the 90-day limit on most regular discount loans to depository institutions under section 10B of the Federal Reserve Act. Because the loans were extended under 13(3), details on who used the facility are being released to the public now after a one-year lag versus the two-year lag for regular discount window loans.

Use of the program

The Federal Reserve created the program to reduce the strains in the banking system that developed when SVB and Signature Bank failed, sparking runs on some other banks.^[2] After the program opened, lending rose sharply as banks used the loans to replace lost deposits or simply to bolster liquidity reserves to prepare for possible withdrawals.

Borrowing rose sharply again starting at the end of November, when changes in interest rates made borrowing an arbitrage opportunity. As noted, the loan rate was the one-year OIS swap rate plus 10 basis points. Moreover, the loans could be repaid without penalty. Consequently, if the one-year OIS swap rate fell to more than 10 basis points below the interest on reserve balance rate (the "IORB" rate, the rate the Fed pays on deposits), a bank could take out a loan, leave the proceeds on deposit at the Fed and earn the interest rate spread. If the IORB rate subsequently declined to below the loan rate, the bank could immediately repay the loan. As shown in Exhibit 2, the loan rate dipped below the IORB rate occasionally in the early weeks of operation, and it then moved continuously below on Nov. 6, 2023, sparking a substantial increase in borrowing. To eliminate the arbitrage opportunity, on Jan. 24, 2024, the Fed changed the loan rate so that it would never fall below the IORB rate (link).

Table 1 provides statistics on the banks that borrowed and shows how those statistics compare to those for banks that did not borrow. Except where noted, the variable shown is the average of the quintile (1 to 5) of the particular statistic, so the average would be 3 in each column if there were no difference between borrowers and non-borrowers. As shown in the table, borrowers had higher shares of uninsured deposits, greater amounts of unrealized losses on their securities, and experienced larger deposit outflows, all results consistent with banks that were under pressure turning to the BTFP for liquidity support. Borrowers were also more likely than non-borrowers to have borrowed from the discount window during 2022, the last year for which borrowing data are available, suggesting that those borrowers who perceived less stigma associated with borrowing regular discount window credit were more likely to use the BTFP.

Table 1. Statistics on Banks that Borrowed from the BTFP

Note: The sample includes all active commercial banks as of December 31, 2022. Call report data is as of December 31, 2022, unless otherwise noted.

¹ Ratio of uninsured deposits to total deposits.

² Ratio of unrealized losses (gains) on the available-for-sale (AFS) and held-to-maturity (HTM) securities held in the banking book to total assets.

³ Ratio of the change in the amount of uninsured deposits between December 31, 2022, and June 30, 2023, to the amount of uninsured deposits as of December 31, 2022.

⁴ Ratio of the change in the amount of total deposits between December 31, 2022, and June 30, 2023, to the amount of total deposits as of December 31, 2022.

⁵ Ratio of the number of commercial banks that borrowed from the discount window (DW) one or more times in 2022 to the total number of commercial banks.

To better evaluate the significance of these differences and how they affected decisions to borrow, table 2 provides the results from a probit regression analyzing the decision to borrow.^[3] The regression results indicate that financial pressure and willingness to use the discount window were statistically significant contributors to the decision to borrow.

Table 2. Probit Regression Results for BTFP Borrowing

Note: The sample includes all commercial banks. Call report data is as of December 31, 2022, unless otherwise noted. Regionals are defined as non-GSIBs with more than $50 billion in assets. Other variables are defined in the note to table 1.

* = 10%, ** = 5%, *** = 1%

The results of probit regressions (which are designed to explain binary choice variables) can be difficult to interpret. To help, we mapped the results into a probability of borrowing for banks with specific characteristics. About 28 percent of commercial banks borrowed.

Table 3. Probability of Borrowing for banks with selected characteristics

Note: Small Banks are defined as banks with less than or equal to $50 billion in assets. A bank under maximum stress is defined as a bank with AFS+HTM Losses Ratio, Uninsured Deposits Ratio and Uninsured Deposits Runoff Rate all in the fifth quintile.

In sum, banks under actual or prospective liquidity pressure were somewhat more likely to borrow from the facility, as would be expected. Notably, even for a regional bank under maximum pressure, greater willingness to use the discount window in the past boosted the likelihood the bank would use the BTFP by 17 percent. For smaller banks, greater willingness to use the discount window in the past boosted the likelihood the bank would use the BTFP by 19 percent. As we noted in "A Major Limit on the Fed's Crisis Toolkit: Shame" reducing the stigma associated with the discount window would make central bank lending a more powerful tool to address financial strains.

First Republic Bank

The third and final bank to fail in spring 2023, First Republic Bank, borrowed heavily from the BTFP.^[4] In March, FRB borrowed $13.9 billion backed by collateral with par value of $13.9 billion. If the market value of the collateral was 20 percent lower than par, $2.8 billion of the loan was uncollateralized.

First Republic failed and its assets were sold by the FDIC on April 28, 2023. As some detective work by Yale's Steven Kelly revealed (link), at that point the Fed reclassified First Republic's loan to "other loans" on its weekly balance sheet. The FDIC became the obligor of the loan, and the collateral was replaced by a promise by the FDIC to repay it. Because the original borrower had defaulted, the Fed charged the FDIC an additional 100 bps for the loan. First Republic's $14 billion loan was repaid by the FDIC on Nov. 30, 2023. None of these changes in loan type, loan rate, or obligor were reflected in the loan-level data that the Fed released on March 12, 2025.

As noted, when the FDIC took over First Republic's loan, it sold the collateral backing the loan, replacing the collateral with its own promise to repay. Even though the collateral was worth less than the loan, probably about $11 billion, the FDIC repaid the entire amount of $14 billion. The remaining $3 billion was covered by the Deposit Insurance Fund, which is being replenished by the banking system. Although the terms of the guarantee are unknown, it is plausible the FDIC could have decided not to assume the First Republic loan, in which case the Fed could have foreclosed on the collateral and Treasury would have paid the Fed the remaining amount due under its guarantee.^[5] Thus, by assuming an obligation to repay $13.9 billion and collateral worth only about $11 billion, the FDIC converted an obligation of the Treasury to one borne by the banking system. The FDIC has provided no explanation for its action.

Only par value disclosed

As amended by the Dodd-Frank Act, Section 13(3) of the Federal Reserve Act requires that "the security for emergency loans [be] sufficient to protect taxpayers from losses" and requires the Fed to publish, along with other loan data, the "value of the collateral" backing the loan. To assess the risk to taxpayers, it is necessary to know the fair/market value of the collateral backing the loan, not the par value, and the Fed's past practice has been to report fair value minus a haircut. Indeed, the Fed is required by law to assign a lendable value to collateral "consistent with sound risk management practices and to ensure protection for the taxpayer."

While par value determined the potential maximum size of the loan, it is not the right measure for assessing risk. Although the Fed could hold a security with fair value below par until maturity and receive the par amount, along the way the Fed would receive below-market interest payments. The cost to taxpayers is the same: The realized loss if the Fed were to sell the security immediately and the negative carry if it were to hold onto the security have the same present value.

Notably, in the Fed's first required letter to Congress on the BTFP, on March 16, 2023 (link), the Fed listed the credit protection provided by Treasury as one of the features of the program "intended to mitigate risk to the Federal Reserve and taxpayers." [emphasis added]. Of course, credit protection from the Treasury does not mitigate risk for taxpayers. By the second letter a month later, the Fed no longer claimed that the Treasury backstop protected taxpayers, just that it protected the Fed (link).

The chart below shows estimates of the uncollateralized exposures that the Treasury was exposed to over the life of the life of the BTFP. In the absence of disclosure of uncollateralized amounts by the Fed, we have assumed that market values of the collateral averaged 80 percent of the par values.^[6] Notably, these estimates indicate that the aggregate uncollateralized exposures never exceeded $25 billion, suggesting that the Fed was adequately protected from losses by the credit support from Treasury.^[7]

Conclusion

The BTFP accomplished its objective of calming fears that banks would not be able to meet depositor withdrawals, but at significant cost. One bank, First Republic, failed with a $14 billion loan outstanding backed by approximately only $11 billion in collateral. Moreover, loans were extended at a below-market rate against insufficient collateral. Central bank best practice when providing emergency credit is to lend at a high rate so that borrowers are encouraged to repay promptly when they can. Lending against abundant collateral protects taxpayers from risk and prevents the central bank from engaging in fiscal policy. To be sure, the FDIC's decision to repay First Republic's BTFP loan after it failed protected taxpayers, albeit only indirectly, by effectively shifting the losses from the under-collateralized loan to the banking system via the Deposit Insurance Fund.

The program may have had mixed effects on the significant stigma associated with borrowing from the Federal Reserve. On the one hand, the attractive terms led many banks to borrow, and more frequent borrowing reduces stigma. On the other hand, the low rate and under-collateralization could add to the public view that discount window loans are bailouts, a significant source of stigma. The Fed could have prevented the BTFP from becoming an arbitrage opportunity rather than strictly a source of liquidity support to address withdrawals by preventing the lending rate from falling below the IORB rate from the start.

The Fed may have chosen to extend undercollateralized loans at a below-market rate for one-year terms so that bank depositors were reassured that those banks under stress would have the funds needed to meet withdrawals. To avoid having to take on such risk in the future, the Fed should encourage banks to have abundant collateral at the discount window and provide greater assurance that the Fed will extend regular discount window loans to any solvent bank with adequate collateral. It could do so by recognizing the discount window borrowing capacity in liquidity regulations and strengthening its commitment to lend.

[1] As discussed below, the Federal Reserve Act requires that the security for each loan made under a 13(3) program is sufficient to protect taxpayers from losses.

[2] A New York Fed staff study "Tracing Bank Runs in Real Time" found that 22 banks experienced runs.

[3] A probit regression is a statistical method for estimating the probability of a binary outcome such as borrow/don't borrow.

[4] SVB Bank and Signature Bank failed before the BTFP was launched.

[5] The precise details of the Treasury guarantee have not been disclosed, but the Treasury provided $25 billion in credit guarantees using the exchange stabilization fund.

[6] A 10-year Treasury note purchased before the rise in interest rates in 2022 would have had a market value equal to about 75 percent of its par value in the spring of 2023.

[7] In the absence of clarity regarding the specific terms of the credit support the Treasury provided the Fed, one cannot be certain.