Essentials of Risk Management (3rd edition)

Introduction

Silicon Valley Bank’s (SVB) collapse on March 10, 2023, was the second-biggest bank failure in U.S. history after Washington Mutual Bank in September 2008. According to the Federal Reserve, the bank was the 16th largest in the U.S., with $209 billion of assets as of Dec. 31, 2022. SVB was an outlier due to its highly concentrated business model serving the technology and venture capital sector, a high percentage of uninsured operating deposits, and a significantlong-term investment in rate-sensitive securities.

SVB was flooded with cash during the pandemic tech boom. SVB used a lot of that money to buy Treasury bonds and mortgage-backed securities. But as interest rates rose, those securities declined in value. That was not a problem at first as long as their depositsremained stable since SVB could ignore, from an accounting standpoint, any losses from a declining value of the assets. Tech tumbled after the Federal Reserve began raising rates aggressively in 2022 to curb inflation, from close to zero in March 2022 to around 5% in March 2023. As a result, startups withdrew their deposits faster than the bank expected. In early March 2023, the extent of deposit withdrawals was such that SVB was forced to sell $21 billion of bonds and realized $1.8 billion in market losses.

​SVB’s deposit base depended on the financial performance and behavior of a concentrated list of VC-backed technology companies.SVB’s $161 billion domestic deposits were divided into $82 billion interest-bearing and $79 billion non-interest-bearing deposits. In addition, SVB ranked first among banks with more than $50 billion in assets, with approximately 95% of its total deposits being uninsured. Also, SVB invested a significant amount of its $117 billion domestic deposits in long-dated government and agency-issued mortgage back securities, of which $26 billion were classified as Available For Sale (AFS), and $91 billion were classified as Hold-to-Maturity (HTM) . As of year-end December 2022, the total HTM securities portfolio had a weighted average duration of 6.2 years, with maturities of 10 years or more. SVB’s HTM portfolio as a percentage of total securities was nearly double the average for Large Banking Organizations (LBO) and its uninsured deposits as a percentage of total deposits were more than double the LBO average of VC-backed companies.

NII Risk Calculation

SVB management believed the one-year Net Interest Income (NII) risk calculation was asset sensitive based on the wrong assumptions. The NII risk calculation did not fully reflect the idiosyncratic risk associated with the uniqueness of its customer base and how rising interest rates could negatively impact the bank.This was partly due to the short-term nature of the uninsured deposits on the liability side since these deposits could be rapidly withdrawn. On the positive side, there was a mix of short-term floating-rate loans ad fixed-rate sensitivities on the asset side. For example, the maturity of $64 billion of the $74 billion of loans and leases held for investment was less than one year as of year-end 2022.

The NII risk calculation in the banking book was based on parallel shifts in the yield curve. It showed that NII was exposed to falling rates (see Table 1) and did not reflect the massive withdrawal of deposits that eventually occurred. Earnings at risk (EaR) or net interest income (NII) at risk captures short-term exposure to interest rate movements. It measures NII volatility generally over a one-year horizon based on yield curve shocks. For example, firms will shock interest rates by 100, 200, or more basis points (bps) in either direction and then estimate the impact to NII. Various yield curve shocks and twists can be used for this exercise. Deposit assumptions are important for this analysis as firms must assume the amount of the market rate movement they will pass through to deposit accounts (also known as “deposit betas”).

SVB did not publicly disclose potential longer-term negative impacts to earnings highlighted by an Economic Value of Equity (EVE) metric at year-end 2022 but publicly disclosed their EVE metric as of year-end 2021. EVE estimates the structural mismatches of a bank balance sheet relative to yield curve movements. It is viewed as a longer-term measure since it is a discounted cash flow approach that estimates the present value (PV) of balance sheet cashflows to calculate the EVE (EVE=PV of assets – PV of liabilities). Changes in EVE are done by shocking interest rates by various amounts (e.g., +/− 100, 200, or more bps) to estimate exposures as cashflow paths change. Deposit assumptions are important in this exercise, so cashflows must be estimated based on customer characteristics. 

​SVB’s failure to publicly publish the EVE for year-end 2002 was a red flag. Moreover, the Risk Appetite Statement (RAS) of SVB set by the board only included the NII risk calculation, not the EVE calculation. In addition, SVB disclosed the NII risk calculation to the asset-liability committee (ALCO) based on a down 100 basis points and a 12-month ramp-up in interest rates instead of a large range of potential plausible shocks and did not provide ALCO with an EVE calculation.

March 8th through March 10th ,2023

As previously mentioned, the rise in interest rates, such as Fed funds rate from near zero in March 2022 to 5% in March 2023, negatively impacted the value of long-dated securities and resulted in pressure on earnings and potential losses. SVB sold $21 billion of AFS securities, primarily U.S. Treasuries, on March 8th,2023, and incurred a $1.8 billion loss. SVB also announced a plan to launch a $1.75 billion share sale to shore up its balance sheet on the evening of the 8th, but share prices fell 60% on the 9th. SVB required the share sale to plug the $1.8 billion loss caused by the AFS.

The AFS losses triggered an accelerated deposit outflow at SVB. Social networks, media, and other ties reinforced the accelerated run dynamic. On March 9th, deposit outflows were over $40 billion, and SVB projected an expected $100 billion deposit outflow more the next day. The automation of new technological advances in managing deposits enabled clients to remove deposits, quickly creating systemic consequences through contagion. The liquidity risk turned into a solvency crisis. Shares of SVB fell another 60% in premarket trading on the 10th before being halted and never reopened for trading. This led the California Department of Financial Protection and Innovation (DFPI) to close the bank on March 10th. 

Model Risk

The interest rate and liquidity risk measures used by SVB contained significant model risk. As we discussed, SVB failed to document the impact of a longer-term interest rate risk behavior such as provided by an EVE measure. Further, SVB could have made risk disclosuresmore transparent for stakeholders than was evidenced on the surface by accounting disclosures, such as performing integrated reverse stress tests that would have shown the potential vulnerabilities in the rapidly growing amount of unrealized losses in the AFS and HTM security portfolios. SVB’s models also failed to capture idiosyncratic liquidity risk in stress markets, such as the accelerated and massive withdrawal of uninsured deposits. The challenge is that no bank can have sufficient liquidity to survive if customers massively withdraw their deposits at the scale and pace that occurred as SVB.

Regulatory Considerations

Regulatory scrutiny for SVB increased when they crossed the $100 billion-dollar threshold in average total consolidated assets and met the criteria for a category IV bank under the 2019 tailoring rule and became subject to Enhanced Prudential Standards (EPS) under regulation YY. SVB failed the Internal Liquidity Stress test (ILST) and its contingent funding plans were weak because , among other weaknesses, it  did not have a sound plan for converting its assetsinto cash .The lack of a sound plan should have led to immediate action to remedy the breach of regulation YY, but that did not occur.  

            Regulators are now exploring the need to put in place a stronger risk management framework to correct deficiencies observed in the SVB case. For example, the Liquidity Coverage Ratio (LCR) and Net Stable Funding Ratio (NSFR) did not apply to banks below the $250 billion-dollar asset threshold. As discussed in Chapter 3 of EoRM3, the Basel Committee set global minimum liquidity standards to make banks more resilient to potential short-term disruptions in access to funding and addresses longer-term structural liquidity mismatches in their balance sheets. The LCR requires that banks maintain unencumbered high-quality liquid assets (HQLA) sufficient to withstand net cash outflows over the next 30 calendar days based on a stressed funding scenario specified by the regulator. The LCR funding scenario includes a combination of idiosyncratic (institution-specific) and market-wide shocks. Theshocks include a run-off of retail deposits and an increase in market volatilities that, in turn, affects the quality of collateral and margin calls. If SVB had been subject to the LCR requirement, SVB would have had a 9 percent shortfall in HQLA in December 2022. An estimate for February 2023 showed an even more significant shortfall in HQLA of approximately 17%.

Aftermath

Following the failure of SVB, doubts on the soundness of other midsize regional banks spread among uninsured depositors who started to withdraw their deposits. Signature Bank and First Republic Bank were closed by the regulators a few days after the fall of SVB. To avoid further contagion the government took the decision of guaranteeing the full amount of insured and uninsured deposits. Also, the Federal Reserve offered a program to make loans secured against long-term Treasuries and mortgage-back securities without a haircut on the market value of the securities offered as a collateral. There was the fear that more than 200 banks would be at risk of failure if half of uninsured depositors withdrew their money from the banks. 

​We can expect additional regulation to correct the deficiencies uncovered in the SVB case. For example, introduce additional capital requirements against unrealized losses in the AFS account.