Let’s ask the CEO of the bank: “what will bring you down?” and, “how will you respond?” I suggest other governance and reporting fixes as well.
This question related to how reporting should change post SVB and FRB gets asked quite a few times in my class. More recently, my friend Ethan Heisler, the editor of The Bank Treasury Newsletter, asked me to address this question for a CFA (Chartered Financial Associates) Society panel that he is scheduled to chair.
I suggest that banks report a liquidation analysis under scenarios related to the risk factors that they disclose in their own 10-Ks. Let me explain.
What is liquidation analysis?
Inspired by my work with Venkat Peddireddy of CEIBS, I believe that the following thought experiment run by an outside analyst might be useful:
· Estimate the callable liabilities by considering all the contractual obligations (on- or off-balance sheet) that are likely to be called at the onset of some liquidation event that can cause a run on the bank. What might such an event look like? This is where scenario analysis and risk factors in the 10-K come in. In SVB’s case, the liquidation event turned out to be a steep rise in interest rates. We will come back to the risk factors that SVB discussed in its final 10-K a bit later in this piece.
· Identify the assets on the balance sheet that can be sold to pay off the callable liabilities. We assume that intangible assets are worthless in a liquidation.
· Estimate the potential haircuts that the bank will incur upon the required liquidation of assets. Charge the losses from sales of these assets and potential asset write-downs (unrealized losses) caused by the liquidation event against the firm’s book value of equity.
· Estimate the post-crisis haircuts for the remaining unsold assets by applying an appropriate loss default rate to each asset according to its credit rating or its historical loss rates during a crisis.
· Finally, compute capital short falls as the excess of all the above haircuts over the bank’s book value of equity.
How would this work for SVB?
Callable liabilities
Start with the schedule of “contractual commitments” disclosed in on page 86 of SVB’s 2022 10-K. For the uninitiated, this schedule reports the payments that the bank is expected to contractually make in aggregate. That aggregate number is broken into less than one year, 1-3 years, 4-5 years and after 5 years.
The aggregate contractual obligations number, as per SVB’s 2022 10-K is $193.2 billion. Of this, $189 billion is due within a year. Of that number, $173 billion relates to deposits and the remaining $16 billion relates mostly to short term borrowings.
How much of this $173 billion deposit base will become callable or run in the event of a crisis? That is difficult for us to know from the outside. However, I can at least refine this number a tiny bit by throwing out insured deposits, which I assume will not run when a crisis happens. It turns out that SVB’s insured deposit base was small. On page 80 of the 10-K, SVB reports that their uninsured deposits were $151 billion out of $173 billion.
As a first cut, assume all the $151 billion and the $16 billion of borrowings, or a total of $167 billion, will potentially run or become callable when a crisis hits. Closer examination of the data suggests that $13 billion of the $16 billion comes from the FHLB (the Federal Home Loan Board). Given that the FHLB’s job is to support mortgage lending, one can perhaps assume that this $13 billion will not become callable. Hence, let’s assume that the callable liabilities work out to $154 billion (151+16-13).
What about the possibility that SVB could borrow more from the FHLB? I assume that a last-minute loan from the FHLB, when a bank run is on, is difficult. Indeed, the Wall Street Journal article “How the Last-Ditch Effort to Save Silicon Valley Bank Failed,” suggests that the day before SVB failed it had sought a $20 billion loan from the FHLB of San Francisco, but the FHLB was only willing to provide a smaller amount, perhaps because it did not have the cash on hand and would need to raise it.
Hence, I have ignored the possibility that the FHLB could have been tapped at the last minute for the calculations to follow.
Which assets can I liquidate?
Let’s turn to the assets side of SVB’s balance sheet and start with the most liquid assets at hand. SVB reports $13 billion of cash, available for sale (AFS) securities of $26 billion (reported at fair value) and $91 billion of the infamous held to maturity (HTM) securities.
Will the sale of AFS securities realize the stated fair value on the financial statements? The 10-K discloses that the portfolio of $26 billion consists primarily of $16 billion of US treasuries, $6 billion of agency MBS (mortgage-backed securities) and roughly $1.4 billion of agency backed CMBS (collateralized mortgage-backed securities). “Agency” in this context refers to Freddie Mae and Fannie Mac, the so-called government sponsored entities (GSEs). Because agency securities are, in effect, backstopped by the US government in one way or the other, we expect the haircuts on the sale of these securities to be minimal.
In the 2008 crisis, haircuts associated with prime rated US agency securities was 1% and US treasuries was 0.5%. Applying these haircuts to AFS portfolio yields losses of $154 million ($80 million for US treasuries and $74 million for agency MBS and CMBS securities).
Let’s turn to HTM securities, which primarily represent US agency backed securities. As is well known, the $91 billion represents cost of acquisition, as opposed to the fair value of these securities. Page 164 of the 2022 10-K states that the estimated fair value of HTM securities is $76 billion. This suggests that the remaining $15 billion is loss of value in HTM securities that needs to be charged to book capital if these securities had to be sold on December 31, 2022. On top of that, the sale will likely result in 1% haircut as stated before, which works out to $76 million.
As of now, we have realized roughly $115 billion of cash ($13 billion of cash + $26 billion of AFS and $76 billion of HTM securities) minus the haircuts to pay off the $154 billion of runnable liabilities. Where does the remaining $39 odd billion come from? And what kind of non-temporary write downs in loans and other assets will the crisis entail?
At this point, we have run out of relatively liquid assets to sell. So, we must negotiate the sale of the loan portfolio to some other bank. What kind of haircuts should these loans take?
SVB’s loan portfolio as of December 31, 2022, amounted to $74 billion. The three major categories of loans include (i) around $41 billion was devoted to what they call “global banking”; (ii) $10 billion to private banking loans; and (iii) $8 billion to something they call “innovation” based loans.
One option is to consider the average loan loss rates under the severely adverse scenario, published by the Federal Reserve Bank, every year since 2013. That portfolio loss rate for loans works out to 6.2%. If we were to assume that 6.2% of the $74 billion loan portfolio were to be written off, we are looking at $4.6 billion of a charge to equity. The remaining realization of around $69 billion on the loan portfolio can easily settle the $39 billion cash hole that we were trying to fill earlier.
Charges to book equity
Thus far, we have discussed a $15 billion fall in value in the HTM portfolio and a $4.6 billion charge on account of the loan portfolio. The AFS and HTM haircuts add up to $230 million. The resultant $19.8 billion loss will more than wipe out the $16.2 billion of equity reported as of December 31, 2022, by SVB.
A subtle but important point: it is important to go through an “as if” liquidation scenario and come out as solvent. If a bank fails this “as if” thought experiment, the chances of an actual deposit run, and the possibility of real insolvency, go up manifold.
How is our measure different from LCR and NSNR?
As an aside, I looked up the so-called “living will” that SVB filed with the FDIC (Federal Deposit Insurance Corporation) to check how much of their planned resolution, should a crisis, occur, resembles the plan proposed above. I was disappointed by how sketchy the resolution report looked. It relies heavily on committed but unused credit lines from the FHLB and the Federal Reserve. As reported, these credit lines somehow failed SVB when the real crisis came.
One may also wonder whatever happened to the liquidity capital ratio (LCR) requirements, which require banks to hold a portfolio of high-quality liquid assets at least as large as expected total net cash outflows over a 30-day stress period. SVB states that LCR requirements do not apply to them. Yale published an as-if exercise of applying the LCR rule to SVB and concluded that SVB would have fallen short. How is the LCR different from what I suggest above? LCR addresses liquidity over 30 days and is quite prescriptive in how the actual liquidity is calculated. That is, a regulator prescribes what portion of the deposits will run, for instance, and how much of the agency MBS securities will be sold. Instead, I ask the CEO to estimate these parameters based on the private information they have about the bank. The other conceptual difference is that our process combined liquidity concerns with solvency. LCR focuses only on liquidity.
How is our measure different from NSFR (net stable funding ratio)? The rule requires banks to have a minimum amount of stable funding backing their assets over a one-year horizon. The rule defines funding as stable, based on how likely it is to be available in a stressed environment and classifies assets by type, counterparty, and time to maturity. Yale goes through another as-if analysis and concludes that SVB would have passed the NSFR test. As noted in that Yale report, regulator-mandated formulae don’t factor in idiosyncratic information the CEO of the bank has. For instance, the rule assumes that 50% of SVB’s deposits are “available” although that may not be the case given highly concentrated deposits held by VC backed clients of SVBs. And the NSFR rule effectively ignores unrealized losses on agency MBS securities.
How would our measure work for First Republic Bank (FRB)?
If one were to apply the same analysis to First Republic, we do not end up with a capital deficit. Why? FRB’s latest available 10-K suggests that $119 billion of deposits were uninsured. Even after accounting for appropriate haircuts on its $32 billion of AFS and HTM securities and on its loans of $166 billion, the charge-offs will not exhaust the $17 billion of book equity. How is FRB different from SVB? FRB held a far smaller portfolio of HTM securities relative to SVB. And, FRB had a far larger portfolio of insured deposits, which we have assumed do not run.
So, why was FRB in trouble? Potentially because of the contagion from the SVB incident. As suggested later, future FRBs could report an “as if” liquidation analysis based on the scenario related to the risk that contagion from other weaker banks causes our own depositors to flee.
What’s the way forward?
I suggest that banks conduct and publicly disclose an “as if” liquidation analysis, of the kind I just described, assuming scenarios relevant to their risk profiles. An outside analyst’s estimates of what amounts are realizable, based on publicly disclosed data, are necessarily coarse. The bank’s Chief Risk Officer (CRO) has or should have far more fine-grained information about potential concentration of deposits, the chances that these deposits would run especially in today’s social media age, the realizability of liquid and illiquid assets should a liquidation event occur. On top of the other unknowns mentioned in my simplified “as if” liquidation, there are several significant details related to the “as-if” liquidation, that I, as outsider, don’t know much about:
· Which assets and liabilities, especially derivatives linked to specific assets and liabilities, are implicitly matched but not reported as offsetting assets and liabilities?
· As suggested by my colleague, Doron Nissim, does the presence of intangible assets, such as a well-known brand or a customer interface technology system, reduce the probability of a liquidation or potentially increase what a prospective buyer would pay for the liquidating bank?
· Will only 40% of the uninsured deposits run when a stress event happens, as assumed by the LCR methodology, relative to the 100% we assume, as suggested by my colleagues, Kairong Xiao and Suresh Sunderasan?
How would CEOs know the answers to these questions?
Well, if they or their risk officer does not know, then we have a bigger problem. Alla Gil, the CEO of Straterix, an economic scenario consulting firm, suggests that CROs project forward macroeconomic and market scenarios, where stable conditions are interrupted with shocks and model how the ripple effects change the correlations between risk drivers as well as behavioral patterns of depositors, obligors, investors and the like.
What would a stress event look like? How are these different from the stress tests that the Fed runs?
I suggest that banks customize the definition of a liquidation event to the risk factors they disclose in their 10-K. These risk factors may or may not overlap with the macro-economic events that the Fed tends to emphasize such as rise in unemployment rates, interest rates and the like. And, of course, the Fed stress tests are applied to banks above a size threshold. As is well known by now, SVB fell just below the Fed specified threshold for mandatory stress tests. My hope is that all banks will report scenario based “as if” liquidation analyses.
For completeness, let’s take a closer look at the risk factors that SVB discusses in its last 10-K. It turns out that SVB lists a grand total of 32 risk factors. If you are cynical, you could argue all these risk factors are legal boilerplate disclosures and not diagnostic. Perhaps. I want to believe there is some kind of signal in these risk factors. And, if they are indeed uninformative boilerplate, we need to make sure they are diagnostic of the real risks facing the bank, going forward. Which of these is more likely to lead to a liquidation event? I don’t know and I hope management does. I would have potentially picked the following six risks as more pressing:
· Interest rate spread declines: “Our interest rate spread may decline further in the future. Any material reduction in our interest rate spread could have a material adverse effect on our business, results of operations or financial condition.”
· “Liquidity risk could impair our ability to fund operations and jeopardize our financial condition.”
· Condition of markets: “Our equity warrant assets, venture capital and private equity fund investments and direct equity investment portfolio gains and losses depend upon the performance of our portfolio investments and the general condition of the public and private equity and M&A markets which are uncertain and may vary materially by period.”
· Systemic risk: “The soundness of other financial institutions could adversely affect us.”
· Deposit growth: “We have experienced significant growth during 2021 and into 2022, including deposit growth. If we again experience deposit growth at a similar or greater rate than has occurred in the past, we may need to raise additional equity to support our capital ratios.”
· “Concentration of risk increases the potential for significant losses, while the establishment of limits to mitigate concentration risk increases the potential for lower revenues and slower growth.”
Of course, this may be totally off-base. May be the next crisis comes from a cyber-attack or a sovereign default or something else. But these risks should hopefully show up in the risk factor disclosures. And the CEO or the risk officer has some sense for which of these tail risks is likely to be serious and whether and how the bank can respond if one of these tail risks were to materialize.
What about other accounting and disclosure fixes?
I have deliberately stayed away from asking for other fixes such as charging HTM unrealized gains and losses to book value of the equity. Historically, the FASB (Financial Accounting and Standards Board) has struggled with the proper role of fair value in financial statements of banks. Part of the issue has been the reliability of market data and the potential for manipulation compared to historical cost, as was the case leading up to the Great Crash in 1929.
But we saw fair value become front and center in the 1970s and 1980s as banks failed and their underwater bond portfolios caused huge losses for the Federal Deposit Insurance Corporation (FDIC) as the receiver. The pushback from industry is that fair value accounting is essentially lopsided, given that it is not generally applied to the liability side of a bank’s balance sheet. Another problem with fair value is that it only applies to the bond portfolio, not to the loan portfolio. Hence, fair value ends up creating confusion and accounting compromises such as AFS.
My accounting and finance colleagues who are steeped in banking research themselves cannot seem to agree with what needs to be done with fair valuing the balance sheet, for instance. For instance, Kim, Kim and Ryan (2022) suggest eliminating (i) the currently allowed classification of debt investment securities as held-to-maturity (HTM); and (ii) the regulatory accumulative other comprehensive income (AOCI ) filter. These proposals would require banks to recognize the effects of unrealized gains and losses on debt investment securities for financial reporting and capital purposes. Other colleagues such as Trevor Harris suggest that public financial reporting is too discrete to be meaningful and “fair value” can mean different things in dynamic and volatile markets for tradable securities let alone OTC securities or loans.
A couple of other modest requests include:
· Depositor and loan concentration: Perhaps banks can publicly disclose the top 10 depositors and loan relationships with the underlying dollar numbers to help us understand concentration of deposits and loans.
· Better derivatives disclosure: NSFR is essentially a disclosure about how much of the bank’s long-term assets are funded with short term liabilities. Central to this claim is the extent to which interest rate risk is unhedged. Jiang, Matvos, Piskorski and Seru (2022) suggest that only 6% of aggregate assets in the U.S. banking system are hedged by interest rate swaps. On top of that, the derivatives disclosures on banks’ financial statements are currently too opaque to answer the question of the extent to which this funding gap is hedged via derivatives. Improvements in clarity in derivatives disclosures would help.
The ”hedge effectiveness” test imposed by accounting rules are arguably too restrictive to accommodate macro hedging of the balance sheet, as opposed to micro hedging of a derivative to a specific item on the balance sheet or the income statement. Is it time to think about accommodating macro hedging in the “hedge effectiveness” framework?
· Risk management experience: The presence of a risk officer in the C-suite and a board member with risk management experience in the banking context can be a big red flag. Perhaps regulators can mandate the presence of a risk officer in the C-suite and a board member with risk management experience.
· Compensation: SVB’s CEO was paid on the bank’s financial performance without much regard for the risk associated with earning that performance. My colleague, Suresh Sunderasan, suggests that CEOs should ideally be paid the way hedge fund managers are: using a Sharpe ratio approach, meaning, financial performance scaled by some measure of the volatility or the risk associated with such performance.
Having said all this, my call for scenario based as if liquidation analyses is simple and intuitive. The approach is based on the risk factors unique to a bank, regardless of size or other thresholds, and gets out of the business of prescribing top-down fixes. Instead, we ask the CEO of the bank: “what will bring you down?” and, “how will you respond?”
Read the full article here
