Skip to main content

The 2023 Banking Crisis: Lessons about Bail-in

The FDIC could have saved $13.6 billion in the recent failures of three large banks if regulators had earlier held the banks to the total loss-absorbing capacity (TLAC) standard, based on reasonable assumptions.

The FDIC could have saved $13.6 billion in the recent failures of three large banks if regulators had earlier held the banks to the total loss-absorbing capacity (TLAC) standard, based on reasonable assumptions.

Over the past few months, US regulators resolved three large, failed banks: Silicon Valley Bank (SVB), Signature Bank, and First Republic Bank. The FDIC estimated that coverage of uninsured depositors, discounts on the purchase price of the banks, and potential losses on the assets it assumed would cost its Deposit Insurance Fund (DIF) a total of $31.5 billion, roughly half of which it will recoup later from other large banks.

We estimate that the FDIC could have saved $13.6 billion by imposing losses on other creditors of the failed banks if US regulators had held the banks to a modified version of an existing international standard known as total loss-absorbing capacity, or “TLAC,” based on reasonable assumptions (Figure 1).

Under the 2016 TLAC standard agreed by the Basel Committee on Banking Supervision, regulators began requiring the world’s largest banks to maintain a minimum level of long-term, unsecured debt to complement their regulatory capital. Together, this debt and regulatory capital (equity) represent a bank’s TLAC. If a bank’s losses exhaust its capital, the additional layer of debt can be “bailed-in,” that is, converted to equity, providing an additional cushion against the bank’s insolvency.

The presence of TLAC facilitates the open-bank resolution of bank holding companies. Regulators now expect the largest bank holding companies to have resolution plans under which they can enter bankruptcy without shutting their banking and investment banking subsidiaries. TLAC complements these plans because it recapitalizes these subsidiaries and keep them open.

Currently, US regulators require all banks to meet minimum capital requirements. But they only require eight global systemically important banks (G-SIBs) to meet the TLAC requirement, which includes the layer of long-term debt on top of capital. In late 2022, US regulators proposed extending the rule to the five or so additional domestic banks with greater than $250 billion in assets.

In the wake of recent bank failures—all three of which were below this $250 billion threshold—the Biden Administration renewed calls to apply the long-term debt requirement to more US banks. On a recent earnings call, PNC Financial CEO Bill Demchak said, “TLAC, I think, is a certainty at this point” for his $550 billion bank, yet uncertainty remains about “how much it will be and whether it’s varied as a function of size and complexity of a bank.” Although regulators have not yet proposed a new threshold, FDIC Chair Martin Gruenberg noted at the time of the 2022 proposal that the agency was considering the challenges posed by banks with $50 to $100 billion in assets. In 2019, Paul Tucker, former Deputy Governor of the Bank of England, and former FDIC Chair Sheila Bair submitted a letter to Jerome Powell and FDIC Chair Jelena McWilliams, encouraging them to impose long-term debt (LTD) requirements on all US regional banks. There is now growing speculation that regulators could seek to apply the rule to the 20 or so domestic banks with more than $100 billion in assets.

The regulators’ proposal last year did not provide details about how they might adjust the TLAC rule to smaller banks. The recent bank resolutions offer an opportunity to study how a revised rule might have worked if it were already in place.

Background: Basel standards, international variations, and size-based tailoring

Under the Basel TLAC standard, G-SIBs are required to hold an additional cushion on top of their minimum regulatory capital requirements. In addition to common equity, TLAC includes unsecured debt with a remaining maturity of more than one year, so long as that debt is subordinated to deposits and other excluded liabilities. As of 2022, a G-SIB’s total TLAC under the Basel standard should be the greater of (1) 18% of its risk-weighted assets (RWAs)—that is, the bank’s loans, securities, and other assets, weighted to reflect the risk of loss to the bank—or (2) 6.75% of its leverage exposures—that is, the bank’s total assets and off-balance sheet exposures.

Beyond these minimum standards, implementation varies by country. The US applies a minimum 18% of RWA and 7.5% of leverage exposures—again, though, only to G-SIBs. The Basel standard also sets an expectation that LTD will form at least one-third of the TLAC requirement (see page 3 of the standard). Following Basel, the US requires its G-SIBs to have 6% of RWA in LTD, or one-third of the 18% requirement.

Unlike the US, the EU requires all banks to meet a minimum requirement for their own funds and eligible liabilities, or “MREL,” and regulators vary this requirement by the size of the bank. With some minor exceptions, liabilities qualifying for MREL are the same as TLAC: long-term, unsecured debt with at least a year of residual maturity. The primary difference is where those liabilities fall in the creditor hierarchy. For example, unlike TLAC, MREL includes some liabilities, such as senior preferred bonds, that rank pari passu with protected liabilities, such as common claims.[1]

As of 2022, European G-SIBs have to hold a minimum level of MREL at the Basel standard: 18% of RWAs or 6.75% of leverage exposures. So-called “top-tier banks”—non-G-SIB banks with assets exceeding €100 billion and other designated banks—have to hold 13.5% of RWAs or 5% of leverage exposures in MREL, roughly 75% the level required of G-SIBs.[2] EU regulators also have flexibility and discretion to apply an MREL standard based on a bank’s size and financial condition, meaning most G-SIBs and top-tier banks have institution-specific requirements

For this paper, we consider the question: How much more of the losses could the failed banks’ non-depositor creditors have borne, and how much could the FDIC thus have saved, if US regulators had held them to a TLAC requirement?

Of course, it’s impossible to know what requirement regulators would have imposed on these banks if they had imposed one. We could simply assume that regulators would have applied the same rule that they applied to G-SIBs: TLAC equal to at least 18% of risk-weighted assets or 7.5% of leverage exposures. But we think it’s more likely regulators would have taken a tailored approach, as they did in other areas, and applied a standard similar to the one Europeans apply.[3]

Ultimately, the precise level of required TLAC should reflect its purpose, and non-G-SIBs would probably hold TLAC for a different purpose than G-SIBs. The G-SIBs typically have multiple banking and investment banking subsidiaries that would need to be recapitalized—by converting TLAC into equity—in order to stay open through the holding company’s bankruptcy.

In contrast, a non-GSIB’s TLAC might be used to facilitate the FDIC’s most common form of resolution, known as a purchase-and-assumption transaction. In purchase-and-assumption transactions, the FDIC arranges for a bigger bank to purchase the bank and assume most of its liabilities, leaving the remaining assets and liabilities with the FDIC. This is the approach it took in the recent bank failures, except that it first had to create two bridge banks before it could line up buyers for SVB and Signature.

In the cases this year, crucially, the FDIC did impose losses on existing shareholders and creditors of the three banks. That meant total losses for the three banks’ shareholders (about $43 billion, based on separate calculations of the book value and market value at year-end), preferred equity holders ($7.3 billion), and other creditors ($4.7 billion). The total bail-in for shareholders and creditors was therefore about $55 billion (see Figure 2).

The FDIC’s bank-funded DIF was also depleted by an estimated $31.5 billion. Roughly half, or $15.8 billion, resulted from the FDIC’s exceptional coverage of uninsured depositors in SVB and Signature. The FDIC is obliged to recoup the costs of protecting uninsured depositors through special assessments, which it plans to levy on 113 banks. The remaining half, or $15.7 billion, was a loss borne by the DIF due to the discounts offered to the purchasers of the failed banks (not including any loss-sharing agreements with the buyers of the failed banks) and the estimated losses on the assets the FDIC assumed and will eventually liquidate. With a TLAC requirement there would have been more creditors available to absorb losses, providing a first line of defense for the DIF.

This distinction between using TLAC to keep banks open and using it solely to protect the insurance fund is apparent in the European Banking Authority’s most recent report on the MREL capacity of European banks. The largest European banks plan for an open-bank resolution, in which TLAC converts to equity and the bank stays open; most smaller banks plan for a transfer to an acquiring bank, in which TLAC is available to absorb losses (European Banking Authority, page 9).


Based on the above reasoning, we sought to calculate whether Basel’s existing TLAC rules, as tailored by European regulators, could have saved the FDIC’s DIF money by allowing it to bail-in more long-term debt (LTD) creditors. In short, our analysis shows that the three failed banks did not have sufficient LTD at the level or composition that regulators would require for an effective bail-in.

Not all LTD can effectively absorb losses; following the Financial Stability Board’s 2015 recommendations, regulators require any TLAC debt to be unsecured and subordinated to depositors and other protected creditors. Seeking a simple and straightforward bail-in, US regulators in their implementation of the Basel standard went a step further by requiring debt instruments to be “plain vanilla,” that is, without convertibility or derivative-linked features, such as call options. EU regulators, on the other hand, are more generous in what qualifies as TLAC, accepting more Tier 2 instruments, structured notes, and other convertible regulatory capital instruments, which the US does not accept.[4]

Of course, TLAC would be one requirement of many already applied by regulators. For example, these three banks were also subject to a capital conservation buffer of 2.5% of their RWAs, which they had to meet entirely with common equity Tier 1 capital (CET1). If they breached their minimum capital requirements plus this buffer (i.e., a total CET1 ratio of 7.0%), regulators would force the banks to pause capital distributions. To ensure that TLAC is unencumbered, banks cannot count CET1 for both their required buffers and their TLAC requirement. (This is true of G-SIBs, as well, which are subject to more and larger buffers.) For our calculation, we earmark 2.5% of the banks’ RWAs to meet this buffer, meaning it cannot count toward TLAC.

Our analysis applies the EU’s current threshold for similarly sized banks, which is 13.5% of risk weighted assets.[5] We also apply the US rules for eligible instruments and debt-equity composition. That is, we expect an LTD target of 4.5% of RWAs, or one-third of the 13.5% TLAC target, and an equity target of 9% of RWAs.


Based on our calculations, these three US banks would have had a combined equity and LTD shortfall of $13.6 billion under these assumptions:

Silicon Valley Bank. At year-end 2022, Silicon Valley Bank (SVB)’s liabilities included enough TLAC-eligible equity and debt instruments to exceed the EU’s minimum 13.5% threshold. However, those instruments were mostly CET1 or preferred stock, and long-term debt made up just 3% of RWAs. To meet that part of the US standards, SVB would have had to raise an additional $1.7 billion in LTD or convert some of its excess capital into LTD.

Signature Bank. At the same point in time, Signature’s TLAC ratio was only 7.1%, substantively all of which was in CET1. To comply with a 13.5% TLAC threshold and a 4.5% LTD threshold, Signature would have had to raise another $3.5 billion in LTD. On the capital side, we assume Signature would set aside enough CET1 to continue its capital distributions. Excluding the CET1 that would have met the capital conservation buffer (2.5% of RWA) would have placed Signature in a TLAC-capital deficit of $2.3 billion. The combined shortfall was $5.8 billion.

First Republic. First Republic had only $779 million in long-term senior notes outstanding, and a TLAC ratio of 9.5% at year-end 2022. To meet our hypothetical requirements, First Republic would have had to raise an additional $6.1 billion in LTD.


We conclude that, had the three banks raised sufficient LTD and equity to meet a reasonable version of the TLAC standard, those investors would have borne more of the losses—potentially saving the DIF about $13.6 billion, and much more under the standard regulators now apply to US G-SIBs (Figure 1).

Meanwhile, our understanding of the actual write-downs performed during the FDIC-brokered purchase-and-assumption transactions suggests that the fundamental logic of TLAC is sound: The FDIC forced losses on the bondholders and shareholders of these three banks, where it may once have protected them in a “systemic risk exception” (SRE) event. Perhaps more interesting is that the FDIC performed these bail-ins both under an SRE designation (SVB and Signature) and not (First Republic).

In 2019, when Martin Gruenberg was an FDIC board member but not the chair, he said that in a resolution without a buyer, and little or no unsecured debt, the least cost resolution would require uninsured depositors to take losses. It’s reasonable, then, to view the FDIC’s ex ante least cost option for resolution as requiring a bank’s investors to bear more of the losses, protecting uninsured depositors. Our analysis shows that, unsurprisingly, these banks had not issued enough LTD to bail-in, and the extent of their TLAC shortfalls provide insight into how regulators may consider applying TLAC requirements to smaller US banks.

In analyzing their LTD shortfall, we make four observations.

First, in most US banking groups, the top-tier parent company issues all of the long-term unsecured debt, and all deposits and other short-term unsecured debt are issued by the operating subsidiary. Therefore, TLAC-eligible senior debt must be issued by a bank holding company, which structurally subordinates it to the debt issued by the holding company’s subsidiaries. This “clean holding company requirement” allows a resolution authority to bail-in the holding company’s debt so that the operating subsidiaries may continue as going concerns. First Republic and Signature, however, did not operate a bank holding company structure like SVB, technically rendering all of their outstanding debt instruments ineligible for TLAC. If regulators apply LTD requirements to more banks, they must consider whether the banks will be required to establish a BHC structure that issues LTD. This is a question of how easy it is to resolve the entity, since by law bail-in-able debt is statutorily subordinated to protected liabilities.

Second, due to these organizational differences, any TLAC requirement should complement resolution strategies. As we’ve noted above, G-SIBs need TLAC to replenish their capital, such that the operating subsidiaries may stay open while the holding company goes through bankruptcy. Smaller banks need enough TLAC to execute their resolution strategies under Dodd Frank’s Title I, which typically includes the sale of the business to a larger buyer. In response to proposals to apply TLAC to more banks, First Republic stated that its non-BHC structure meant that it did “not present the risks and complexities with respect to financial stability or resolvability” (page 4). To be sure, it was sold faster than the other two banks. But it also had less than $1 billion in LTD to bail-in and cost the FDIC’s DIF an estimated $13 billion.

Third, a bail-in is supposed to prevent losses to insured depositors, and in this respect, there is no fundamental difference between holding the loss-absorbing capacity in debt or equity.[6]

However, regulators will need to consider the proper debt-equity ratio in any tailored TLAC requirement. The five largest US banks not currently subject to the TLAC requirement recommended that regulators allow them to raise either debt or equity (Comment Letter, page 23). We expect US regulators would set a debt target at one-third the TLAC target, like the FSB recommends, and perhaps for good reason—in principle, debtholders are more attuned than equity investors to downside risk.[7] The existence of both debt and equity also helps regulators clearly distinguish between going concern capital and gone concern LTD. Of course, a bank with more equity should enjoy lower costs of debt funding, too.

Fourth, ensuring a minimum amount of TLAC and LTD requires continuous monitoring by banks and their supervisors. For example, one month before SVB’s fateful press release, First Republic redeemed $500 million of its senior debt, roughly two-thirds of its outstanding TLAC-able debt. The current US rules prevent G-SIBs from redeeming LTD if doing so would cause them to breach their TLAC or LTD requirements. In Europe, redemption of any MREL-eligible liability requires the resolution authority’s approval and supervisory consultation.[8] The ex ante cost of this maintenance and supervision often seems onerous until we see the ex post cost of paying out depositors and managing contagion.

A fundamental power of resolution authorities like the FDIC is the power to subordinate the claims of long-term unsecured debt to those of short-term unsecured debt. This ability to retrospectively discriminate among creditors is designed to reduce the run incentive of depositors and other short-term debt holders.

More importantly, the resolutions of these three banks demonstrate that the FDIC is willing to bail-in shareholders and creditors. Nevertheless, the DIF will bear losses above this bail-in to protect uninsured depositors. These losses are later socialized through banks via special assessments.

The question for the regulators and the regulated will thus boil down to a difficult policy choice between increasing self-insurance through TLAC or increasing socialized insurance through the DIF. Bankers tend to prefer the latter: TLAC is a certain cost; and that debt and equity funding is considerably more expensive than deposit funding, and probably more expensive than any future special assessments. Regulators may lean more on TLAC, particularly for the larger banks, to avoid the risk of depleting the DIF and rendering the FDIC less able to cope with further contagion during times of financial stress.

The authors wish to thank Jordan Thursby, Paolo Colonna, and Stephanie Quiet for their helpful input and comments. The views and analysis are those of the authors and remain personal to them.

Figure 1: TLAC Deficit Calculations
Figure 1: TLAC Deficit Calculations

Figure 2: Costs of Resolution to DIF and Bail-in Investors
Figure 2: Costs of Resolution to DIF and Bail-in Investors

[1] The EU, unlike the US, does not apply a minimum debt ratio to its TLAC (see page 7 of this comparison). However, it does set an additional subordination requirement, called a bank’s total liabilities and own funds (TLOF), which is set at 8% of RWAs.

[2] European authorities also designate some banks with less than €100 billion in assets as other systemically important institutions, which are subject to the lower MREL standard. Furthermore, European banks face additional MREL requirements due to capital buffers and supervisory add-ons; in practice, the average MREL requirement for large European non-G-SIBs was about 20% of RWA at the end of 2022, according to the Single Resolution Board (page 5). Practically all European banks will meet the EU’s capital conservation buffer, at 2.5%, if not an entire combined buffer requirement (CBR), which includes the capital conservation buffer, a countercyclical buffer, a systemic risk buffer, and a systemically important bank buffer.

[3] Our calculations do not apply a TLOF-like standard, not only because these three banks had different corporate structures, but for the more essential point that US law statutorily subordinates a bank’s other liabilities to its depositors. US regulators only impose subordination requirements on intermediate holding companies, that is, the US holding companies of the US operations of foreign banks, which are not bound by US statutory subordination.

[4] There are many differences in the US and EU rules. For example, regulators in the US and EU only count LTD with a residual maturity of at least one year, but the US goes further by haircutting eligible LTD by 50% if it matures in less than two years. This is a material difference in rulemaking, but for the purposes of our analysis, we leave this issue aside since all the eligible LTD issued by these three banks had remaining maturities of greater than two years.

[5] We do not use the leverage exposure measure since these banks, unlike G-SIBs, are not subject to consistent leverage reporting requirements.

[6] From the bank’s perspective, TLAC-eligible debt is generally more expensive than other types of debt, but cheaper than equity. However, the composition of debt and equity changes the bank’s incentives. More equity may reduce risk-shifting behavior, but it may also reduce returns, and thus encourage risk-taking. More debt may discipline bank management since it would incentivize monitoring by creditors, and therefore may prevent the bank’s insolvency long before a bail in would be necessary, but it also affects the leverage ratio of these banks. Because preferred equity counts toward capital requirements without diluting shareholders, regulators should carefully consider how bail-in-able these instruments are.

[7] Debt is generally easier to price, and price volatility for a bank’s debt may offer regulators a market-based signal for directing supervisory resources.

[8] For banks engaged in market making, the European rules grant renewable, year-long exceptions to this prior permission mechanism.