Comment on the New G30 Report

The Group of Thirty released a new report on Tuesday, “Bank Failures and Contagion; Lender of Last Resort, Liquidity, and Risk Management”.  The report is available here.  The Group of Thirty is an organization of past and present leaders of central banks and other financial agencies around the world.  The working group that wrote the report was chaired by Bill Dudley.  Stijn Claessens was the project director and Darrell Duffie and Trish Mosser were project advisors. The report consists of an analysis of the spring 2023 failure of Silicon Valley Bank, Signature Bank, First Republic Bank, and Credit Suisse and a series of related policy recommendations.  The key recommendation of the report is that all banks maintain collateral at the discount window plus reserve balances equal to uninsured deposits and short-term borrowing, and, in support, that the Federal Reserve improve the efficiency of its collateral management operations. 

We commend the G30 and authors for a thoughtful report and support many (though not all) of its policy recommendations.  The report recognizes that central bank lending is a critical part of bank liquidity risk management and needs to be factored into regulatory requirements and supervisory expectations for bank liquidity, that it is better for society if banks keep making loans to businesses and households that they can pledge as collateral to the discount window than simply requiring banks to hold ever-larger stockpiles of reserves at the Fed or lend even more to the U.S. Treasury, and that the entire liquidity regulation framework warrants careful review and reconsideration.  At the same time, the report emphasizes that banks bear primary responsibility for their liquidity risk management and should not be protected from poor decisions. 

The report also recognizes that the implementation of the key proposal will require a lot of additional analysis about available collateral and the stability of different types of bank liabilities, and that some banks have specialized business models that may require adjustments.  Although a lot of work needs to be done, the report’s main recommendation provides a good starting point for careful policy study of the future of the bank liquidity framework and the role of central bank lending in that framework.

Recommendations With Which We Disagree

Tailoring.  The report recommends that the existing tailoring of regulations by bank size should be reduced because the events in spring 2023 demonstrated that the failure of mid-sized banks can be systemic.  While we agree that the failures of the certain mid-sized banks had systemic consequences requiring a strong government response, those consequences were a function of circumstances, not those banks’ size, interconnectedness, or role in the economy.  In particular, SVB and Signature failed largely because they were unprepared to access the discount window for emergency funding, and similarly situated banks were also perceived as insufficiently prepared and thus at risk of run.  Consequently, if the report’s recommendation on discount window preparedness is adopted, mid-sized banks would not pose the types of systemic risks that arose when SVB and Signature failed.  In addition, a key reason why the failures of SVB and Signature had systemic consequences was the uncertainty, delay, and disruptions associated with the resolution of these banks by the FDIC, yet steps to improve the FDIC’s capacity to handle resolutions quickly and more effectively is a needed reform not mentioned in the G30 report.

Frequency and intensity of examinationsThe report also recommends that the banking agencies conduct more frequent supervisory reviews of banks and escalate supervisory concerns for penalty or enforcement more often and more rapidly.  But as we have described at length elsewhere, the supervisory problems that contributed to the banking turmoil of spring 2023 was not too infrequent examination or too patient supervisory, but rather a misguided supervisory approach that was principally directed at issues that had little to do core safety and soundness concerns, a consistent focus on process over substance, a failure to apply rules already on the books, and a reliance on supervisory ratings that depended on subjective judgments and not objective data.[1]  The report thus misses the mark on needed supervisory reforms; rather than supervisors being ever more aggressive, they should be more focused on issues that actually present a risk to the bank.

There are a few areas where the report may have left the reader with the wrong impression.

  1. The report states that the accounting treatment of held-to-maturity securities is inconsistent with regulatory treatment of those securities, which count as high-quality liquid assets that may be used to satisfy liquidity requirements.  HTM securities are recorded on bank’s books at amortized cost rather than market value.  It would of course be inappropriate for HTM securities to be included in HQLA at amortized cost, but they are not; rather, they are included in HQLA at market value.  Moreover, the accounting rules governing HTM securities permit them to be pledged as collateral in repo funding markets, pledged to the discount window, or pledged as collateral at the Fed’s standing repo facility, therefore making them legitimate and reliable sources of liquidity. 
  2. When proposing that more resources should be devoted to supervision, the report references the “fiscal costs” of bank failures.  There were no fiscal costs—no cost to U.S. taxpayers whatsoever—from the bank failures in spring 2023.  All costs incurred by the FDIC in resolving SVB and Signature, including payments to insured and insured depositors, were and will continue to be paid for by the banking industry alone in the form of Deposit Insurance Fund assessments.
  3. When arguing that the LCR’s 30-day outflow assumptions on deposits should be increased in light of the rapid outflows observed this spring, the report states that the current outflow assumption for nonoperational corporate deposits is 40 percent.  However, that is the assumption for nonfinancial corporations.  For such deposits for financial firms (such as SVB’s deposits from venture capital funds), the LCR’s outflow assumption is 100 percent.  These types of deposits are frequently a significant portion of everall nonoperational corporate deposits; for example, the 2023Q3 LCR disclosure of a large bank that also works primarily with financial institutions reports assumed outflows under the LCR for nonoperational wholesale deposits in the aggregate equal to 96 percent.  While regulatory calibrations should always be reviewed in light of new experience, the LCR’s current deposit outflow assumptions are much more conservative than the report suggests.
  4. The authors note that if their recommendations are adopted, the LCR will become “largely irrelevant” in stress times, but that “[t]he LCR and the related HQLA requirements remain relevant for bank’s liquidity management in normal times.”[2]  But the LCR is a stress metric –that is, it is expressly designed to reflect liquidity needs and resources of banks under an acute period of liquidity stress – and thus not at all appropriate as a guide for liquidity management in normal times.  As one example, the LCR assumes the bank can get no liquidity from other banks in the interbank market nor even withdraw cash it has on deposit at another bank, ignoring the usual central role of the interbank market under normal liquidity conditions. 
  5. The report strongly supports reducing the severe stigma associated with borrowing from the discount window.  However, it states that banks should use the discount window only during periods of stress, and not in normal times, and that the pricing of discount window loans be made more attractive.  This may not be a successful strategy for reducing stigma.  Prior to 2003, the U.S. discount rate was at or below market rates, and banks were told they could use it when it was appropriate to do so, but not too much.  This was substantially changes in 2003, when the Federal Reserve raised the rate so that its rules on use could be eliminated; the precisely aim of these changes was to reduce stigma by allowing banks to use the window whenever they wanted, with the increased price of discount window loans alone serving as a limit use of the window.  In some ways, the report appears to be recommending a return to the pre-2003 regime.
  6. The report states that the AOCI filter should be removed for banks with more than $50 billion in assets as a way to improve the “…interaction between the current [mark-to-market]/HTM accounting (and supervisory) rules and their impacts on capital and earnings.”[3]  However, the AOCI filter only changes the treatment of gains and losses on available-for-sale securities not held-to-maturity securities.

Conclusion

Well-designed regulations are rooted in reality.  As demonstrated in spring 2023, a bank that is prepared to use the discount window is more liquid than a bank that is not, and when all banks are prepared, the banking system is more resilient.  We support what appears to be the core message of the G30 report:  that regulatory and supervisory assessments of individual banks’ liquidity should recognize the important role played in stress times by the lender of last resort as a source of funding.  Changing regulations takes time, but the Federal Reserve has already shifted in this direction by reminding the public and its supervisors that the discount window is an important component of banks’ contingency funding, and reaffirming that use of the window should not be viewed negatively.  Going forward, we agree that a fundamental reconsideration of liquidity requirements is in order.


[1] Jeremy Newell and Pat Parkinson, “A Failure of (Self-) Examination:  A Thorough Review of SVB’s Exam Reports Yields Conclusions Very Different From Those in the Fed’s Self Assessment,” May 8, 2023.  https://bpi.com/a-failure-of-self-examination-a-thorough-review-of-svbs-exam-reports-yields-conclusions-very-different-from-those-in-the-feds-self-assessment/

[2] P. 28

[3] P. 21