Silicon Valley Bank was the 16^th largest bank in the US, and its collapse caused a lot of turmoil in the financial market worldwide. No bank can survive if all of its depositors demand funds at the same time. No bank keeps this much liquidity; otherwise, there is no incentive to conduct business. The actual default occurred due to a classic bank run. So, it’s essential to understand what led to the loss of confidence for SVB depositors.

I believe six key issues led to the collapse of Silicon Valley Bank, and I have explained them below –

1. Challenging Macro Environment

Since the great financial crisis of 2008, the Federal Reserve in the US has been running an easy money policy called Quantitative Easing. Interest rates were near zero, and it was super easy to get funding for anything. The table below shows the growth of the Fed balance sheet –

Except for the brief period from 9/17-8/19, the balance sheet size consistently grew as Fed pumped liquidity into the economy. Especially during the Covid 19 pandemic, the balance sheet ballooned and doubled from its pre-pandemic levels. Easy monetary policy was coupled with the unprecedented fiscal policy by the US Government. Due to the pandemic lockdowns, economic activity was low, and this stimulus money found its way into bank accounts. Banks across the US were flooded with new deposits and didn’t know what to do with them. Many banks invested these deposits into safe US treasuries and Agency MBS securities to generate some yield.

The Fed underestimated the impact of this easy monetary policy on inflation (maybe because the pre-pandemic policy did not result in higher inflation in the US). However, the pace and magnitude of the COVID-era policy were very high, and in hindsight, maybe the Fed should have paid more attention and started raising rates sooner. In Feb 2022, Russia attacked Ukraine, which further complicated the inflation issue. The global supply chains were impacted, resulting in higher oil prices and an inventory shortage. All this resulted in higher inflation in the US, but the Fed initially deemed it transitory, which was incorrect.

As the Fed got its inflation forecast wrong, it was left to catch up on the monetary policy front, resulting in the fastest interest rate increases since the 1980s. I explained later that this complicated balance sheet management for banks and SVB was caught on the wrong side of the trade. In the below graph, one can see the steepness of the current rate increase cycle, also called Quantitative Tightening –

The Fed is committed to fighting inflation and may raise interest rates even further if required. Raising rates has evaporated liquidity from the financial markets and reduced people’s purchasing power, resulting in pressure on businesses. Due to this tightening, venture capital funding dried up, and the pre-revenue startups and tech firms were left tapping into their deposits to fund their day-to-day operations, which led to deposit outflows for a bank like SVB, which had concentrated exposure to this client base.

Another consequence of higher rates was the drop in bond prices. When rates go up, the bonds issued when rates were low fall in price as new bond issues pay higher interest rates. So, when buyers have the option to invest in securities yielding higher coupon payments, the low-yielding bonds logically drop in price to have a similar return profile. This issue impacted SVB substantially as most of its assets were held in US Treasuries and Mortgage-Backed Securities.

2. Sensitive Deposit Base (Liabilities)

• Industry Concentration –

As the name suggests, SVB was the bank of Silicon Valley. The below stats are from SVBs investor presentation as of December 2022.

Except for the 6% share of the private bank deposits, all other clients are related to the technology and venture capital sector, i.e., 94% deposit concentration in a similar industry. A similar concentration can be observed in its off-balance sheet funding mix. For all such clients, Silicon Valley Bank was a go-to place. SVB marketed itself as a firm with strong expertise in serving these clients, which other traditional banks would not have served with favorable offerings as SVBs as most of these firms tend to be in the pre-revenue stage or focused on the innovation economy, which may take longer to materialize business potential. Below is one more stat from the SVB’s website which highlights the concentration of its client base –

• High Balance of On-Demand Uninsured Deposits –

The startup tech industry was booming during the pandemic, which saw explosive growth in SVB deposits. SVBs deposit base shot up from $61.7bn (Dec 2019) to $189bn (Dec 2021), i.e., a three-fold increase in 2 years. Almost 40% of the deposits were non-interest-bearing (compared to 10-20% for large banks), so they had less incentive to stick with SVB. More than 90% of the deposits were uninsured, as these accounts had higher balances than the FDIC-insured limit of $250k, as these were business deposits that firms use for payroll or working capital needs. However, this also means that these firms were susceptible to any negative news around SVB as they feared losing all funds over $250k, which would have been catastrophic for these startups to survive. As per some estimates, the average deposit size at SVB was ~$4 mil compared to the $177k average at smaller regional banks. For example, Roku had ~$500 mil, and Circle (which owns the stablecoin USDC) had ~$3bn deposited with SVB. Most of these firms had no other banking relationships, and their maximum funds were tied to SVB, so any potential issues with SVB could cause panic for these firms as they were afraid of losing all their funds.

• Digital Banking –

The tech-savvy clients of SVB were mainly using services online. SVB had only 16 physical locations, including corporate offices. Even if we assume all 16 locations also acted as a branch, still, it’s a tiny physical footprint for a bank with ~$210bn asset size. It’s also surprising that SVB lacked controls on online wire transfers. Generally, banks have daily withdrawal limits for wire transfers. For example, JP Morgan has a daily limit of $250k for business accounts. SVB customers could put in $42bn worth of withdrawal requests on the same day, i.e., ~20% of total deposits running out of the door on the same day, and SVB systems allowed these requests to go through.

In summary, SVBs had an extremely sensitive deposit base. Deposits were concentrated in one industry, lacking the safety of FDIC insurance, and the SVB lacked controls such as daily withdrawal limits to prevent the sudden surge of withdrawal requests. It was super easy for clients to withdraw all their funds using a cell phone.

3. Mismanagement of the Interest Rate Risk (Assets)

• Substantial MBS Exposure With Duration Risk and Variable Maturities –

As mentioned above, SVB had a 3-fold increase in deposits between Dec 2019 and Dec 2021. As interest rates were near zero during this period, SVB invested these deposits into long-term US Treasury bonds and mortgage-backed securities (MBS) to generate yield. The simple logic is that, in normal times, the longer the maturity, the higher the bond yield. Most of the securities SVB invested in had no or limited credit risk. However, these securities were subjected to massive interest rate or duration risk. As bond prices are inversely proportional to interest rate increases, the market value of SVB’s holdings was at risk of a significant drop in market value.

SVB had a very high proportion invested in MBS compared to other banks on the street. As of Dec 22, out of the $93bn investments, only $7.8bn was cash, $16.2 bn were US Treasuries, and $7.7bn were Gennie Mae issued MBS (which has exclusive US Govt. guarantee); however, rest all ~$60bn was non-Gennie Mae MBS, which was not guaranteed by US Govt. but provided a higher yield. However, this higher yield comes with higher risk. Unlike US Treasuries with a fixed maturity, MBS have variable maturities or negative convexity. This causes MBS prices to fall faster than a typical bond. Check this link for more details on Why Mortgage-Backed Securities Are Negatively Convexed?

In short, material exposure to MBS increased SVBs duration risk. Below is a screenshot from the SVBs investor presentation. Please note that the $117bn balance shown is not marked to the market as a large portion of it is HTM (Held to Maturity) and hence looks higher than the numbers I quoted above. However, you can note the MBS concentration – ~75% of the portfolio was in MBS compared to other banks like Citi or JPM; this ranges from 8% to 12%.

• Lack of Hedging –

SVB had some interest rate swaps to hedge the AFS (Available for Sale) portfolio, but strangely SVB unwound those positions in Q1 and Q2 of 2022. So SVB had a highly volatile MBS portfolio with no hedging.

Maybe SVB did not expect the interest rates to go up so quickly. If a bank classifies securities as held to maturity, it does not need to mark the value down as it intends to hold it until maturity, when it expects to receive the entire principle. So, it does not make sense to hedge the HTM portfolio and incur costs if you hold it till maturity. However, the firm may be forced to sell these HTM securities in case of deposit outflows, which happened with SVB, and it ended up booking losses on the sale. Considering how sensitive the SVBs deposit base was, its management should have done better in terms of hedging. It is also unclear why SVB unwound the swaps, which were at least hedging its AFS portfolio.

4. Less Stringent Regulatory Framework

Generally, firms should not depend on strong regulation and supervisory oversight to better manage risk. Regulatory rules are the same for all firms and may lack firm-specific product-level accurate assumptions. Still, they certainly can help compare all banks with the same rule set and highlight potential anomalies. It’s impossible to know if the bank has conservative internal stress-testing assumptions and risk management processes as this information is not public. Hence, the regulation offers a way to compare banks against the same rules or stress factors. SVB was not subjected to stricter regulations as it was a Category IV bank per the US tailoring rule due to less than $250bn in assets. SVB lobbied with other smaller regional banks to have this tailoring rule passed and have liberal regulations for smaller banks. You can notice that Cat IV banks had concessions on almost all critical metrics like AOCI opt-out for capital rules, lower LCR and NSFR thresholds, Reduced frequency of FR 2052a reporting, and a few other concessions. Similar regulatory requirements (as larger banks) could have highlighted some of these issues sooner to regulators and SVB management; however, it’s impossible to say this for sure.

As SVB had explosive growth, it’s clear that its risk management function could not catch up with the speed of business growth and lacked the sophistication required for a bank of its size. Not being able to hire a CRO for nine months is one example. Moreover, it’s not easy to build risk functions of a bank and have them fully up and running effectively in a year or two to give some benefit of the doubt to SVB. Strong regulation would have helped, especially the short-term liquidity metric called Liquidity Coverage Ratio (LCR).

There is much discussion on HTM (Held to Maturity) classification, and some commentators also referred to it as “Hide to Maturity,” but I disagree that it would have mattered if SVB had a robust internal liquidity stress test (ILST) or subjected to the stricter LCR thresholds. Banks were allowed to not take AOCI (Accumulated other comprehensive income) charges on capital, but liquidity metrics always consider the mark-to-market valuation of assets, and hence, HTM classification wouldn’t have mattered for the ILST or liquidity coverage ratio (LCR) calculation. Also, Capital reporting is quarterly, and a lot can change in a quarter without the public and regulators getting first-hand information. On the other hand, liquidity reporting/metrics are monitored daily (or at least monthly for smaller banks), i.e., ILST, FR 2052a, and LCR.

• Liquidity Coverage Ratio –

US G-SIBs must submit a daily liquidity monitoring report called FR 2052a to the FRB, which has all the data needed to calculate LCR. Based on the daily FR 2052a report, regulators can calculate LCR and question a bank if they see a deterioration in the liquidity position. Due to the tailoring rule, SVB was not subjected to the daily FR 2052a reporting and had lower LCR thresholds. The formula of the LCR is –

Below is the comparison of LCR requirements between a G-SIB and Cat IV Bank –

• The minimum requirement for the US G-SIBs is 100%, whereas, for Cat IV banks, it’s 70%.
• Cat IV banks can further reduce their net outflows to 70% – Refer to the FR 2052a appendix VI for the specifics.

The table below shows the difference in requirements and their impact on a Cat I G-SIB bank vs. a Cat IV bank. To get the same LCR result of 125%, a Cat IV bank can have $114 in outflows compared to just $80 for a Cat I G-SIB. In addition, a Cat IV bank will have a cushion of 55% against minimum requirements compared to 25% for the Cat I bank.

In summary, in current regulations, it was ok for a smaller Cat IV bank like SVB to take more risk but still maintain the same liquidity metric as a large G-SIB like JPM. Note Col D and I. Somehow, a smaller bank with fewer resources and sophistication may take more risk under the current regulatory regime. Therefore, it was easy for a bank like SVB to remain compliant with regulatory requirements, which may have given the management and supervisors false assurance. However, there was one more critical point around LCR compliance. Bank Policy Institute (BPI) and Yale University published an analysis of SVBs LCR. BPI posted an update later. Yale estimated SVB’s LCR to be 75%, and BPI estimates range between 75% to 101%. Both analyses show that SVB would have complied with the current regulatory requirement of a minimum of 70% LCR. First, it’s impossible to forecast LCR accurately based on just the call report or financial statement data. So, I don’t entirely trust these analyses. Any Treasury practitioner would know that US LCR and FR 2052a rules are very detailed and require a lot of interpretation and methodology calibration, especially for the products impacting the denominator of the LCR. A comment in the Yale analysis states that “it’s a myth that LCR is a complex rule.” I disagree with this statement. Yes, the formula of the LCR is super easy, but all the interpretation and methodology from FR 2052a and US LCR rules are incredibly detailed and complicated.

Yale’s analysis concluded that the stricter LCR requirements would have helped SVB manage the risk better. But, on the other hand, BPI concluded that LCR would not have helped SVB manage interest rate risk. So, both conclusions are exactly opposite to each other. Both are somewhat right or somewhat wrong at the same time.

Stricter LCR requirements would have helped, provided SVBs’ interpretation and methodologies of the US LCR guidance were robust (Yale assumed this is the case); however, below, I have explained (next para) that SVB had limitations in interpreting and implementing the US LCR guidance. Stricter requirements may have increased regulatory scrutiny and found some of these interpretation issues (BPI ignored this possibility), but we can’t say that for sure.

Both Yale and BPI analysis missed an important US LCR rule (Page 61526) around HQLA eligibility, which is LRMS (Liquid and Readily Marketable) standard below and other (haircut increase/price decline) requirements per Section 20 (page 61529) in the US LCR. Basel LCR rule does not include LRMS requirements; the US regulators added them in their version. All HQLA-eligible securities, including even level 1 Gennie Mae-issued MBS securities, must meet these requirements. Only US Treasury securities are exempted from the LRMS requirements.

Some banks have developed sophisticated methods to analyze securities or use third-party solutions like Bloomberg Liquidity Assessment (LQA) score as a proxy to meet these requirements. Even though LRMS rules are not specific and are not intended for interest rate risk management, these requirements essentially force banks only to have on-the-run fixed-income securities in their HQLA portfolio. As on-the-run bonds are issued recently (max 3 to 4 months ago), these are automatically not subjected to the severe interest rate risk as off-the-run securities.

Please review this interagency staff report to learn more about the loss of market liquidity, i.e., the increase in the bid-ask spreads of off-the-run bonds (Pages 13, 14 & 20). This report analyzes the US Treasury market activity during market stress, which is higher quality collateral than the MBS securities held by the SVB (It’s clear that if off-the-run US Treasury securities faced market liquidity issues during stress, then MBS will fare even worse).

Both BPI and Yale assumed $25bn of the Level 2A MBS securities held by SVB were HQLAs. This assumption led them to take a $21.2bn value for these MBS in the LCR numerator. Almost all of these off-the-run long-duration MBS securities held by SVB would have failed LRMS and price decline tests, i.e., deemed non-HQLA with $0 credit in the LCR numerator. Now look at the LCR public disclosures (Q4 2022) of JPM below. Note that JPM has absolutely no Level 2A MBS securities in HQLA. Major G-SIBs were discouraged from holding these off the run MBSs due to the stringent LRMS rules, higher regulatory LCR thresholds, and supervisory scrutiny. Even though the LRMS standard is vague, the US LCR rule asks firms to perform security by security analysis to show compliance, which will be very complicated. So, the easy route is not having anything other than level 1 security in the HQLA and ensuring compliance. Note that US Treasury securities are exempt from LRMS requirements. Most banks ignored LRMS requirements, maybe because they were not specific or complicated to implement due to different interpretations, and implementation would have lowered the yield generated on the HQLA portfolio. Based on my experience, most of these off-the-run MBS securities had Bloomberg LQA score of less than 30 (out of 100), i.e., illiquid even if you apply lower LQA threshold criteria and hence can’t be considered HQLA; maybe the reason JPM hasn’t included them in HQLA.

JPMorgan –

Assuming SVB’s US LCR would have been 75% per BPI or Yale analysis, one should adjust that number after reducing most of the $21.2bn MBS. If SVB were subjected to higher standards, like JPM, it might have discouraged them from investing ~70% of their HQLAs in off-the-run Level 2A MBS, which eventually caused massive MTM losses and led to panic. SVB’s LCR and ILST would have started to have a negative impact almost a year before it collapsed with stricter assumptions for off-the-run bonds, forcing SVB to correct its course.

One more critical requirement from the US LCR is that banks periodically test their operational capabilities to monetize the HQLA portfolio. US G-SIBs subject to the full LCR have largely implemented the operational requirements to monetize HQLAs. The larger banks even have monetization plans by asset class for internal stress test purposes ( per Reg YY requirements). These monetization plans estimate monetization capacity from Day 1 to Day 30 by asset class, considering various private (Triparty/Bilateral, etc.) and non-private markets (Central bank pledging, etc.) options. The event at SVB showed that the bank lacked asset monetization capabilities in private and non-private channels.

5. Weak Supervisory Oversight

So why SVB’s supervisors did not promptly act or raise concerns over the years is an obvious question. Based on my experience, there is too much focus on the process, not the actual risk. This is a cultural shift in banking supervision, perfectly explained in this Bloomberg podcast by Lev Menand, an associate law professor at Columbia Law School. I am also posting this podcast below. Please listen specifically from 7 min onwards.

There is no alternative to the SME knowledge of risks in banking and the ability to understand and raise those risks. Just focusing on process, governance, and controls is not enough to find actual risks, and that’s precisely what I believe went wrong with the supervision. There were signs of trouble at SVB as it became the largest borrower of the San Francisco FHLB (Federal Home Loan Bank), but regulators did not act on these signs.

The same issue (too much focus on the process) is impacting the internal supervision of the banks also. Banks today are flooded with so-called controls and governance experts with limited knowledge of financial products, economics, or formal education in finance. If you go to any bank, you will see large teams run by professionals with limited or no experience in the subject matter or the risk they are auditing. They are solely focused on validating procedures and controls. The cultural shift is so substantial that in most cases, when an internal auditor raises a real issue, the first-line team’s immediate pushback to the auditor is to focus on governance and not comment on the actual risk as he is not authorized to challenge risk methodology topic or lack SME knowledge to do so. The assumption is that if due process is followed, somehow, risk will be managed correctly. I agree with the importance of controls and governance, but a far more critical issue is to have SMEs with relevant education and experience in banking. You can see in the case of SVB that the process was followed, and their committees did discuss interest rate risk-related topics, but no one provided an effective challenge to the business or first line. Banks today spend significant resources on governance functions, increasing the inefficiencies as these teams waste the time of first-line groups on unimportant issues and increase the bureaucracy.

6. Poor Communication

SVB was not the only bank facing MTM losses in its HTM securities. If SVB depositors had not panicked, it may have survived. SVB senior management should have done a lot better to avoid such panic. Instead, the SVB CEO just asked his clients not to panic. He even said SVB could pay its depositors except in one scenario. How can one make such a statement and not expect clients to panic even more? In the end, relatively poor communication played a massive role in SVBs’ collapse.

Conclusion

1. SVB operated in a challenging macro environment with a highly concentrated uninsured deposit base and an extensive HTM portfolio, mainly in MBS.
   
   
2. It failed to manage interest rate risk as it assumed rate increases would reverse soon and removed even the minimal hedges they had on the books.
   
   
3. SVBs US LCR rule interpretation and methodology had gaps. Supervision (both internal and regulatory) failed to highlight these gaps. Compliance with key US LCR requirements such as LRMS may have prevented SVB from holding material positions in off-the-run MBS securities, thereby reducing the interest rate risk.
   
   
4. In the end, poor communication and a lack of controls on online withdrawals expedited the deposit run to the extent that no bank could survive.

What Needs to Change –

Below are some things that I think need to change to make the US banking system more robust –

1. The 2018 tailoring rule must go. It doesn’t make sense to allow smaller banks to have lower regulatory standards when they lack sophisticated risk management practices and resources.
   
   
2. Bank supervisors must focus more on actual risks, methodologies, and assumptions than the process. Internal 3^rd line functions in the banks, such as internal audits, must do the same. For example, it’s not enough to just confirm that the stress test or US LCR methodology was reviewed and approved as per the internal governance process. Instead, bank supervisors must review and challenge the actual interpretation and methodology of rules such as LRMS or operational deposit modeling assumptions, etc.
   
   
3. Regulators may consider introducing specific regulations or metrics focused on interest rate risk management. However, clarifying and enforcing existing US LCR rules would prevent these issues rather than introducing new regulations.
   
   
4. There is scope for an increase in deposit outflow rates in the US LCR, but I don’t think that will matter much. Considering the speed with which deposits left SVB, no stress test can have conservative assumptions anywhere near it. As explained above, maintaining the right amount and composition of High-Quality Liquid Assets is the key to avoiding events that lead to deposit runs.
   
   
5. Banks should improve their counterparty sector concentration monitoring, calibrate internal limits, and stress assumptions conservatively.