BPI Chief Economist Bill Nelson Delivers Remarks on Bank Liquidity at MIT Conference

Remarks Delivered at MIT Golub Center for Finance and Policy 10th Annual Conference

Thank you, Paul, and thank you, Debbie, and the other organizers of the conference.  It is an honor to participate, and I anticipate learning a lot.

Debbie asked me to discuss the implications of the bank failures this spring, drawing on my experience as a regulator and as a representative of the regulated, so let me describe my background.  I’m the chief economist at the Bank Policy Institute.  In that role, I oversee the research function at BPI.  BPI is a research-focused trade group representing the 40 largest banks in the United States, including GSIBs and regionals as well as foreign banks with large U.S. operations.  Previously, I spent 24 years at the Federal Reserve Board in the Division of Monetary Affairs where, among other things, I oversaw the Fed’s regular and emergency lending function, and was Monetary Affair’s liaison to Supervision and Regulation.  In that capacity, I was on the committee that ran supervision of the largest banks as well as the steering committee of the horizontal liquidity reviews, and helped design, both in Washington and Basel, the post-global financial crisis regulatory regime, especially the liquidity regulations. 

I’m not going to go into detail about the bank failures this spring, of which I assume you are all familiar.  I will focus on the failure of Silicon Valley Bank, about which I am most familiar, but the failures of Signature and First Republic Banks resulted from similar causes.  SVB incurred large losses on its longer-term fixed-rate assets when interest rates rose sharply over the preceding year.  Uninsured depositors of the bank became worried that they would not be fully repaid and ran.  Because the deposits were concentrated and correlated, the outflows were stunningly fast and massive.

Some of the lessons from the events this spring are not things that we learned, but rather things that we knew but had forgotten.  When I teach my students at Georgetown about banks, I start by telling them that the business of banking – taking deposits and making loans – is inherently unstable, but that the instability has been addressed by deposit insurance, access to the discount window, and regulation and supervision.  But SVB was not protected by the safety net.  Over 95 percent of its deposits were uninsured, it was effectively unable to tap the discount window, and its examiners ignored clear and present risks to safety and soundness and instead focused on process and internal controls. 

One of the lessons is something those of us who study and teach finance know in our bones but have been unable to convince the rest of the world:  Market values, not par values, are what matter; losses are losses whether they are the result of credit defaults or rising interest rates and whether you realize the loss or not.  If you sell the asset on which you made a loss, profits are reduced immediately; if you hold on to the asset, profits are reduced over time; but the present value of the loss is the same either way.  If your plans require your holding onto the asset to maturity, that just means that the liquidity risk is greater.

The link between liquidity and solvency is also being relearned rather than learned.  From 1956 to 1960, the Fed assessed a bank’s capital adequacy using form FR 363.  The form established a risk-based capital requirement that included higher risk weights for assets with credit or interest rate risk.  Moreover, it estimated how much liquidity a bank might need under stress and evaluated the assets it might therefore need to sell at liquidation value.

We also relearned that requiring a bank to hold liquid assets in proportion to its liabilities reduces financial stability and that liquidity support under stress needs to come from the central bank.  Current liquidity requirements, which require banks to hold high-quality liquid assets or “HQLA” in proportion to their runnable liabilities, are essentially the same as reserve requirements.  But the Fed was created because it was understood that reserve requirements made financial stability worse, not better.  If you are just meeting your reserve requirement, then using the reserves to meet a run will result in your falling below the requirement.  In the panic of 1907, for example, many banks stopped meeting redemptions even when they had cash available.  By providing banks discount window credit collateralized by business loans, the Fed made banks more liquid when faced with a run.

There are also several new things that we learned, or at least that I learned.  First, a deposit franchise is not necessarily a good hedge for longer-term fixed-rate assets.  Deposit rates are sluggish, so when interest rates rise, deposit rates fall behind.  If you marked $10 million in deposits with low yields to market, they would be worth less than $10 million, say $8 million.  If the bank owns longer-term Treasury securities with a par value of $10 million whose market value fell to $8 million, the bank would appear to be perfectly hedged, its economic equity would be unchanged.  However, if the depositors decide to withdraw their funds, they get a dollar for each dollar deposited, not 80 cents.  If all the depositors seek to withdraw their funds, the bank’s economic equity declines by $2 million and if insolvent, it would not have sufficient assets to repay all its depositors.  If those deposits are uninsured, the run is essentially inevitable.

Even though some government actions to contain a crisis can have extremely costly longer-run consequences, a government facing a crisis sees the immediate high costs of failing to act.  To prevent itself from taking short-term actions with negative total social benefit, the government has repeatedly adopted laws to tie its own hands.  I learned this spring that the government can get around such restrictions to a greater extent than I had thought.  Because Fed lending to troubled banks in the late 1980s was seen as raising costs to the FDIC, the FDIC Improvement Act of 1991 put limits on the ability of the Fed to lend to troubled banks.  The Fed loans to the bridge banks set up for SVB and Signature Bank seem contrary to the spirit if not the letter of FDICIA. The Dodd-Frank Act required the Fed to only lend under its emergency authority with sufficient security to prevent taxpayer losses.  By providing credit worth more than the underlying collateral, the Bank Term Funding Program seems inconsistent with the spirit if not the letter of Dodd-Frank.  Indeed, the Fed’s first letter to Congress about the lending program states “[t]he BTFP includes features that are intended to mitigate risk to the Federal Reserve and taxpayers,” and points unironically to the credit protection from Treasury as an example. The subsequent letters only state that the program includes features that protect the Fed from losses, not taxpayers.  The Dodd-Frank Act also requires that Congress agree in advance before the FDIC, Fed and Treasury use the systemic risk exception to least cost resolution to provide a blanket guarantee of the deposits of the banking system.  So policymakers, after using the exception to protect the uninsured depositors of SVB and Signature Bank, simply stated that they would be willing to use the exception on a case-by-case basis if other banks failed – no Congressional agreement needed.  Again, the officials seemed to violate the spirit if not the letter of the Act.

There are also conclusions reached by some about the failures that are incorrect.  I read repeatedly, for instance, that the deposit inflows into large banks after the failures occurred because the banks were considered too big to fail.  Well, maybe, but a simpler explanation is that the banks were well known and perceived as safe by depositors who have no interest in studying the financial statements of their bank to determine its risk.  Why point to the prospect of government support, especially after the government prevented losses by the depositors of the smaller SVB and Signature Bank? 

More importantly, contrary to the Barr report but consistent with NYU/Stern’s report issued in July, SVB did not fail because its regulatory requirements had been tailored by Congress and the banking agencies. 

Regarding liquidity, SVB was required to engage in quarterly internal liquidity stress tests, tests that banks report are at least as stringent as the LCR.  Because of tailoring, it was not required to conduct the tests monthly, and it was not required to comply with the LCR or NSFR.  But SVB was awash in liquid assets – 58 percent of its assets were HQLA.  It would have easily passed the NSFR but would have needed $8 billion more HQLA to pass the LCR, a requirement it could easily have met by investing a little more in Ginnie Maes (which are Level 1 HQLA) and a little less in Fannies and Freddies (which are Level 2a HQLA).  Moreover, SVB was consistently failing its ILSTs; the framework was working.  But examiners inexplicably didn’t take action.

Regarding capital, in the absence of the 2019 tailoring rule, SVB’s regulatory capital would have included unrealized gains and losses on its available-for-sale securities, and it would have been subject to supervisory stress tests.  But even after accounting for the losses on its AFS securities, its CET1 ratio would have been 10 percent, well above its 7 percent requirement.  And SVB, with a portfolio that largely consisted of Treasuries and agency MBS, would have easily been able to pass supervisory stress tests because those tests mostly focus on credit losses and assume that interest rates would decline.  Even if interest rates rose in the test, SVB would have performed well, in part because 78 percent of its securities were classified as held-to-maturity and so valued at par in the test regardless of the treatment of its AFS securities.

In light of these lessons learned and relearned, what changes should be made to the regulatory framework and supervisory practices?  On June 9, 2023, the Bank Policy Institute released a set of recommendations that we developed working closely with our member banks.  The recommendations include tightening some aspects of regulation and supervision as well as making material changes to how bank examiners and the Federal Reserve operate. 

In light of the significant role played by securities losses, BPI and its member banks recommended that the set of banks whose regulatory capital includes gains and losses on AFS securities be expanded, that limits be placed on the ability of banks to shift securities out of AFS and into held-to-maturity, and that the Fed include a rising interest rate scenario in its stress tests.  In light of the evident risk of funding a bank with concentrated uninsured deposits, BPI recommended that the banking agencies issue examiner guidance on depositor concentration risk and make any necessary supporting changes to reporting requirements.  Moreover, we recommended that the agencies amend tailoring rules to allow for heightened examination of a bank that grows materially faster than its peers.

We also recommended that the banking agencies conduct a comprehensive review of liquidity requirements and that any recalibration of deposit outflow assumptions be done as part of that review.  We’d called for such a review on February 23, 2023, two weeks before the failures. We pointed out that official assessments of a bank’s liquidity had undergone a sea change following the global financial crisis, and not for the better.  Before the GFC, to be liquid, banks were encouraged to have reliable and well-diversified sources of funding, including being prepared to borrow from the discount window.  After the crisis, attention shifted to requiring banks to hold large amounts of high-quality liquid assets, and the discount window was to be shunned.  As we have been discussing, in the event, SVB failed with a massive amount of HQLA but reliant on concentrated and unstable funding and after struggling unsuccessfully in the final hours to arrange a discount window loan.

In part to address these problems, BPI recommends, and I personally had been recommending inside the Fed since 2010, that liquidity assessments including the LCR and internal liquidity stress tests be augmented to recognize that banks can borrow from the discount window.  What killed SVB (in addition to its losses) was not a lack of HQLA but rather an inability to convert its HQLA into cash fast enough.  As part of internal liquidity stress tests, banks are required to demonstrate that they can monetize their assets, and SVB’s plans were judged by examiners to be deficient.  SVB was actually interested in signing up for the Fed’s new standing repo facility, which would have allowed it to convert its securities into cash, but it was told that its ILST monetization plan could not involve either the discount window or the new standing repo facility.  So SVB did not sign up for the standing repo facility because it would have received no examination benefit.

Our recommendation is similar in spirit to, but more modest than, Mervyn King’s “Pawnbroker for all Seasons” proposal.  King proposes that every bank preposition collateral at the central bank with lendable value equal to all of the bank’s liabilities.  In his plan, neither liquidity nor capital requirements would be necessary.  We are just recommending that discount window borrowing capacity, or some part of borrowing capacity, be recognized as a source of funding under stress, as an amendment to, rather than a replacement for, existing capital and liquidity requirements. 

One way this recognition could be accomplished is already included in the international LCR standard and was mentioned as a possibility in the final U.S. rule that implemented the LCR in 2014.  Specifically, the Fed could sell banks for a fee collateralized committed lines of credit, and banks could count on the unused capacity under the lines as a source of cash inflows in their liquidity assessments.  Doing so would require banks to pay for something they might otherwise get for free, could generate less moral hazard than simply counting existing borrowing capacity, and could help reduce the severe stigma associated with borrowing from the discount window.

Reducing stigma is essential for any plan along these lines.  If banks are not willing to borrow, then borrowing from the Fed should not be recognized as a source of liquidity.  Unfortunately, discount window stigma has been a problem for a hundred years, often because the Fed has encouraged it.  There is no simple solution, but any solution requires that Fed leadership educate the public and Congress that borrowing is a bank’s business decision and should not be viewed negatively. 

I’d end by noting that all the analyses and recommendations that I have discussed this morning have been published by BPI, and anyone interested in any part of my remarks should email me if you want further information.