March 14, 2023
By Oussama A. Nasr*
Background.
Another bank fiasco; another government bailout; another attempt to blame regulators, rating agencies and analysts for failing to anticipate the breathtaking collapse of Silicon Valley Bank (“SVB”). The acronyms HTM, AFS, LCR and NSFR are but four that have become firmly embedded in our daily lexicon in less than a week, and will be explained shortly.
This guide attempts to demystify, for the intelligent layperson, the events that have led to the third largest bank insolvency in US history, while raising a number of questions that have not been addressed elsewhere. We place much of the blame for SVB’s demise on the bank’s risk department and senior management while also holding to account the credit rating agencies, regulators and other third parties for missing several glaring warnings in the bank’s disclosure documents.
By way of background, a bank that acquires a bond must classify it in one of three categories under US GAAP: held-to-maturity (“HTM”), trading, or available for sale (“AFS”). Into the HTM category goes any bond that the bank intends, and has the ability, to hold to maturity. Fluctuations in the price of this bond due to interest rate changes are irrelevant for most reporting purposes; the bank continues to show these bonds on its balance sheet at their amortized cost and ignores any price fluctuations when calculating its income for that period. At the other extreme, in the trading account, all fluctuations in a bond’s value appear immediately in the income statement, and, by extension, on the balance sheet as part of retained earnings. Inclusion of a bond in this category must be accompanied by the bank’s desire to trade this bond in the expectation of making a quick profit. The third and intermediate category, AFS, includes all bonds that are not classified in either the HTM account or the trading account. This category attracts the most complex accounting treatment: fluctuations in value are immediately recognized on the balance sheet, but do not go through the income statement until the bond is sold.
Note that the trading category played no role in the SVB fiasco.
Facts of the SVB case.
The basic facts of the SVB case are not controversial. The bank bought massive amounts of fixed rate bonds when (medium- and long-term) rates where very low and watched their market price (or “fair value”) plunge by billions of dollars during 2022, when those rates rose by some 300 basis points. But having classified these bonds in either the HTM or the AFS categories, the bank was not required to record these losses in its income statement until it sold the bonds, which it refrained from doing until the first half of March 2023. Yet the bank’s published financials as of December 2022 revealed clearly that in regard to the $90 billion or so of HTM securities the fair value had declined by around $15 billion, while for the $28.6 billion of AFS securities the fair value had declined by a further $2.5 billion. Contrast these numbers with the bank’s $16 billion of common equity and $1.5 billion of net income after tax as of that date.
When liquidity pressures grew meaningfully in response to customer deposit withdrawals, the bank resigned itself to the liquidation of “substantially all” its AFS portfolio, thereby realizing an income statement loss of $1.8 billion after tax (or some $2.4 billion pre-tax)1. This loss is broadly consistent with the bank’s disclosure that the bonds in question had a duration of 3.6, since $21bn x 3% (the increase in interest rates since early 2022) x 3.6 = $2.27 billion.
But the bank’s problems did not stop here. The HTM portfolio, it will be remembered, contains another $90bn (in face value) of bonds with an average duration, this time, of 6.2. This causes the fair value of these bonds, given a 3% tightening, to diminish by well north of $15 billion since $90bn x 3% x 6.2 = $16.74 billion2. We reiterate here that these reductions in fair value are not recorded in the income statement unless the bank liquidates the HTM bonds. But with continuing liquidity pressures stemming from deposit withdrawals, the market came to the conclusion that it would not be long before such liquidations became necessary, bringing about the recognition in the income statement of the additional loss of $15 billion – versus the bank’s mere $16 billion of common equity. The rest, as they say, is history.
Questions for discussion.
We address next a number of important questions:
- Should US GAAP be amended to require immediate recognition, through the income statement, of diminutions in an AFS or HTM bond’s fair value? (In practice this would amount to the wholesale abandonment of the current 3-category framework and its replacement with a framework with one or two categories only.) The argument in favor of immediate recognition is that the loss, economically speaking, occurs from the moment interest rates increase, and that this loss is not avoided or reversed merely by refraining from selling the bond. Counterarguments include the fact that the diminution in fair value is disclosed in the notes to the financial statements, and that this permits the reader to assess the true profitability of the institution despite the failure to recognize the loss in the income statement. We sympathize with this counterargument but point out that in the specific case of SVB the market appears to have underestimated the appalling condition of the institution until the liquidation of the AFS portfolio brought about the recognition of the unrealized loss through the income statement. Unrealized mark-to-market losses, somehow, do not “register” in the mind of the reader as vividly as realized losses, no matter how clearly disclosed.
- Even if we continue to allow mark-to-market losses to bypass the income statement until an actual sale of the depreciated bond, should the bank’s capital measures be adjusted (specifically reduced) to reflect these mark-to-market losses immediately? Under this approach SVB’s capital measures would have approached zero during 2022, when the successive interest rate hikes sent the bonds’ fair value plunging by the $15 billion figure (in the case of the HTM bonds) and the $2.4 billion figure (in the case of the AFS bonds). Interestingly, SVB was not required to adjust its capital measure by these MTM losses because it was deemed too small, under US bank regulation, to need to comply with this somewhat onerous standard. Had SVB’s balance sheet assets exceeded $250 billion – rather than amounting only to $212 billion – SVB would have been required to adjust capital in this manner under US bank regulation. Skeptics would argue that the reader is perfectly able to make such adjustments on her own, whether they are required by US bank regulation or not. In other words, the reader could have determined that SVB had become insolvent as soon as the $15 billion and the $2.4 billion MTM losses arose, and well before any liquidation of either portfolio.
- SVB discloses in its 10K that it is not required to comply with either the LCR or the NSFR, again on account of its “modest” size. In a nutshell, the LCR requires a bank to hold “high quality liquid assets” in an amount sufficient to cover the net cash outflow of funds expected in a 30-day period of extreme stress, while the NSFR prohibits the bank from funding illiquid long-duration assets with short-term liabilities. We expect that SVB would have met the requirements of the LCR comfortably, given the plentiful high quality liquid assets spattered on its balance sheet; but we doubt the bank could have met the requirements of the NSFR given the very short tenor of its deposit liabilities. We understand that the exemption of banks like SVB from these two requirements did not become effective until 2018, in response to lobbying efforts by the affected banks vis-à-vis the Trump administration. But whether these requirements are formally required or not, we are surprised that SVB did not choose to apply them voluntarily as part of its internal liquidity risk management.
- A further question on everyone’s mind is the extent to which the holes on SVB’s balance sheet might be replicated elsewhere. The FDIC’s Martin Gruenberg revealed a few days ago that the aggregate unrealized losses across the whole of the US banking universe amount to $620 billion. Separate research by Wall Street analysts indicates that such unrealized losses at J. P. Morgan, Citigroup and Bank of America are material but not life-threatening – except possibly for B of A whose Common Equity Tier-1 capital ratio would plunge from 11.2% to 6% if unrealized losses were immediately reflected in this ratio. To put the $620 billion figure in context, it amounts to roughly four times the common equity of Citigroup, the nation’s third largest bank.
- In its 2022 10K SVB assesses the impact of a 200 basis point increase in rates on its net interest margin, but not on its net worth. Yet for decades, best practices among leading banks have included both such simulations, with results being disclosed, in many cases, by the institution on a voluntary basis. This is particularly relevant at a time when further hikes in interest rates cannot be dismissed, no matter the market’s current expectations. Indeed, SVB would have done everyone a great favor if it had disclosed, back in early 2022 and before the beginning of the interest rate hikes (but after the dramatic expansion of its HTM and AFS portfolios) what the impact of a 200 basis point tightening would be on its net worth.
Conclusion.
All-in-all, a fascinating case study on the worst aspects of inadequate risk management, rating agency underperformance, and regulatory failure. That interest rates would increase rapidly in 2022 was hardly a secret; that fixed-rate bond prices, particularly for long-dated bonds, would decline meaningfully in such an environment was known to the freshest university graduates; and that banks experiencing a run would possibly need to liquidate rapidly a significant part of their investment portfolio was an obvious scenario to contemplate.
*Oussama A. Nasr has worked as a lawyer, banker and financial consultant in New York City and Beirut for 37 years. He began his career with a 4-year stint at the corporate law firm of Shearman & Sterling in NYC, followed by 7 years at Citigroup, before establishing the consultancy DNA Training & Consulting more than 20 years ago. He holds BA and MA degrees from Cambridge University in Mathematics and Philosophy and a Juris Doctor degree from Cornell Law School. His company URL is dnatrainingconsulting.com and he can be reached at onasr@dnatrainingconsulting.com.