Blog Post < Previous | Next >
Guy
How to easily prevent the next mid-sized bank crisis
Ever since the failure of Silicon Valley Bank (SVB) in early March,
there is a sense that the banking industry is a floating iceberg. The
iceberg visible tip represents the 13 largest banks with assets greater
than $250 billion that are fully regulated. The other 4,702 banks make
up the large hidden mass of the iceberg. Thus, we don’t know much about
the financial condition of 99.7% of US banks.
To conduct the following analysis to explore the hidden part of the banking iceberg, I am using a specific bank asset size segmentation defined below.
a>250 = banks with more than $250 billion in assets
b>100 = banks with more than $100 billion & less than $250 billion in assets
c>50 = banks with more than $50 billion & less than $100 billion in assets
d>25 = banks with more than $25 billion & less than $50 billion in assets
e>10 = banks with more than $10 billion & less than $25 billion in assets
f>1 = banks with more than $1 billion & less than $10 billion in assets
g>0.1= banks with more than $100 million & less than $1 billion in assets
h<0.1 = banks with less than $100 million in assets
Banking industry concentration
Now, let's explore the concentration within the banking industry using FDIC data as of 12/31/2022 (the most recent data available at the time of this writing).
As depicted above, the banking industry is very concentrated. The top 13 banks (with more than $250 billion in assets) make up 55% of total banking industry assets and deposits. The top 34 banks (with more than $100 billion in assets) make up 75% of the mentioned assets and deposits. Keep in mind that this level of banking industry concentration is common. It is actually much lower than in Japan, Canada, Australia, Western Europe, and China where typically a half dozen banks dominate the industry.
Within the table below, you can see that the majority of US banks are tiny. 3,725 out of 4,715 banks or 79% of total banks have less than $1 billion in assets. 761 banks have less than $100 million in assets.
The number of small banks (less than $1 billion in assets) has rapidly declined over time. This trend is most likely to continue. As is, the US banking system (even though it is very concentrated) is far more fragmented (with thousands of small banks) than any of its counterparts in the OECD.
What took Silicon Valley Bank (SVB) out?
It was a combination of several factors:
- An inadequate liquidity position;
- Excessive interest rate risk within its securities portfolio;
- A very high level of uninsured deposits;
- A nefarious network effect within its very large uninsured depositors.
To enhance readership and brevity, I will concentrate on the first two items: liquidity and interest rate risk. I have written a longer essay at Medium on the topic, if you care for further analysis and coverage of the current banking industry situation you can access this essay for free at the link below.
Medium essay on the same topic
I am aware that the Media focused on SVB having 93% of its deposits being uninsured as its main vulnerability. For sure this was a material factor in SVB's demise. But, uninsured deposits are not necessarily "hot money." Think of uninsured deposits as primarily business deposits. Such deposits are most often associated with long-standing relationships. Given that a business does not readily change its primary banking relationship. That is because a primary bank often provides so many other related financial services (payroll, treasury management, foreign exchange, letter of credit, loans). Thus, for brevity, I will leave uninsured deposits alone. If you want to read about this specific topic, please read my mentioned Medium essay.
Liquidity
I constructed and tested several different liquidity ratios until settling on a specific one that I call Liquidity Coverage Ratio 1 (LCR 1). The numerator consists of cash-like assets and securities available for sale. They represent assets that are "liquid" and available to cover deposit withdrawals and short-term wholesale funds (banker's acceptance, commercial paper). The denominator consists of uninsured deposits and the mentioned short-term wholesale funds. You can find more technical details on the description of LCR 1 in the Medium essay.
The box plot on the left shows all eight bank asset categories; the one on the right focuses on only the top five bank asset categories (banks > $10 billion in assets or more). Watch out that the legends’ colors are not consistent (I could not figure out how to control the colors efficiently in R).
The box plot on the left shows that the very small banks have very high liquidity. These banks account for only a minority of total bank industry assets and deposits.
The box plot on the right takes the small banks’ noise out of the data. Now, a clearer picture emerges. You can see that the liquidity of the big banks (first box plot from the left) is a lot higher than for any of the following four bank asset categories. The big banks’ 25th percentile (bottom of the box) is very close to or higher than the median for the four lower bank asset categories. This is even the case for the big banks' 1st percentile (around the bottom whisker of the box plot).
Interest rate risk
As a proxy for interest rate risk, I focus on the banks' securities with rate setting greater than 15 years divided by bank assets. These are the securities associated with the most interest rate risk (the longer the maturity, the higher the interest rate risk).
The charts below suggest that, as measured, the big banks do take on more interest rate risk than the others. Yet, this risk seems manageable. Notice that for the big banks, the median securities > 15 years/assets ratio is only around 6%. Yet as shown, this median of 6% is much higher than for the other banks.
There is a lot more to interest rate risk management than just bank securities. It entails the intricate matching of all bank assets and liabilities and the managing of any mismatch with off-balance sheet derivatives. I cover this topic in another Medium article that you can access at the following link for free.
Article on bank interest rate risk management
These interest rate risk granular considerations were way beyond the data I could extract efficiently for each of the 4,715 banks in the US. Nevertheless, as a relative proxy of interest rate risk to compare the different segments of the banking industry interest rate risk level, the ratio I am using is very informative.
How did the combination of interest rate risk and low liquidity take SVB out?
As reviewed a good part of what took SVB down was a combination of inadequate liquidity (low LCR 1) with high interest rate risk (a high ratio of securities > 15 years/assets). Because SVB did not have adequate liquidity, it was forced to sell many of its securities > 15 years. It bought these securities at very low rates during 2021. When rates spiked in 2022 and 2023, SVB's securities > 15 years were much underwater (unrealized losses on these securities were - 16% of par value).
On the chart below, you see how rapidly 30-Year Treasuries rates increased from 1.75% at the end of 2021 to over 3.50% in March 2023 by the time SVB went bust. SVB had loaded its securities portfolio with a lot of long-term Treasuries.
SVB had to sell many of its long-term Treasuries at huge losses to cover rapid uninsured deposit withdrawals. These losses dented SVB's capital base and threatened its solvency. It caused a run on the bank. SVB was done in just a couple of days.
How many SVBs are out there?
That is one of the key questions in our exploring the hidden part of the banking iceberg.
The scatter plot below looking at liquidity on the X-axis and interest rate risk, as defined, on the Y-axis gives us a quick look at where the problem banks are.
The problem banks are the ones located within the red rectangle. They are associated with low liquidity and high interest rate risk, a lethal combination that took out SVB.
So, how many SVBs are out there that we don't know about? It depends on how tight you set your criteria. The table below explores a couple of different scenarios.
The yellow-highlighted portion of the table discloses the liquidity and interest rate risk metrics for SVB. As of 12/31/2022, it had an LCR 1 just under 0.26 or 26% and a securities > 15 yrs/assets ratio just a bit higher than 0.24 or 24%. If we use these same parameters as criteria or filters, we uncover that there are only 7 SVBs out there. There is only one SVB-like bank among banks with more than $100 billion, and it is SVB itself! There is another one among banks with more than $1 billion in assets. And there are 5 banks with just over $100 million in assets. Thus, out of 4,715 banks there are only another 6 banks that are similar to SVB. And, they are all really pretty small banks.
However, if we relax the criteria a bit (LCR 1 < 0.36; securities > 15 yr/assets > 0.14) to uncover banks that are not as bad as SVB, but still have a rather dangerous combination of low liquidity and high interest rate risk, there are now 35 banks. That is still less than 1% of the 4,715 banks out there. And, none of them are among banks with more than $100 billion in assets (except for SVB which is now gone).
The problem with today's bank regulations
The fact that there is currently only about 1% of the banking industry that could be in trouble is not due to sound bank regulations. It is due to sheer luck.
There are really two completely different sets of bank regulations. There is a set for the top 13 banks with more than $250 billion in assets. These regulations are reasonably effective. These big banks do not pop up as problem banks when using the risk metrics we looked at. This is in good part due to the tight regulations they are subject to. However, these regulations are incredibly onerous. They require the very expensive hiring of armies of highly paid quants, database architects, software engineers, and risk managers. These costs are way out of the league of any bank besides the big banks.
The banks with less than $250 billion in assets are dangerously under-regulated. Because the regulations applied to the big banks are deemed way too burdensome for the smaller banks, the latter are not adequately regulated. The smaller or mid-sized banks are not appropriately rmonitored for liquidity and interest rate risk. This regulation leniency is the result of a concerted political effort to support the smaller community banks by not imposing on them burdensome regulations given that they are smaller and not deemed a systemic risk to the financial system.
This logic is devastatingly wrong. This is a perfect way to recreate the S&L crises of the 1980s and 1990s. It took the Government-sponsored Resolution Trust Corporation an entire 6 years (1989 to 1995) to resolve 747 S&L institutions at a cost of $900 billion in today's dollars. That's systemic risk.
One thousand piranhas can be more dangerous than a single great white shark.
How to regulate banks efficiently
The solution is not to under-regulate mid-sized and smaller banks. Instead, it is to fully regulate them, including liquidity and interest rate risk, in an efficient way.
It can be very simple. I am proposing a straightforward dashboard (shown
below) of just five ratios that all banks with no exception would
report to regulators on a monthly basis.
The five ratios would include the following:
A liquidity ratio, LCR 1 (as defined earlier) to equal or exceed 0.40 or 40%. All big banks meet this ratio. Applying this ratio to the entire industry would prevent a prospective liquidity crisis within mid-sized banks and smaller banks.
A capital ratio, Equity/Assets to equal or exceed 0.075 or 7.5%. All big banks, except Charles Schwab, have a higher ratio. Applying this capital standard to all banks would insure they all would be adequately capitalized.
Two interest rate risk ratios.
The first one consists of securities with repricing > 15 years to be equal to or less than 15% of assets. The majority of big banks have a lower ratio. It would prevent all banks from exceedingly loading on long-term securities associated with too much interest rate risk.
The second one consists of Loans with repricing > 5 years to be equal to or less than 30% of assets. This level seems reasonable when observing historical FDIC averages across all bank asset categories.
An asset quality ratio, Reserves to equal 100% or more of all noncurrent loans and REO. This seems like a reasonable level when observing historical FIDC averages across all bank asset categories.
In terms of implementation, I propose that bank regulations adopt atrading clearinghouse approach. In a trading environment, whenever a trader’s margin position falls below par, he or she gets a margin call. Within the proposed bank regulations framework, if a bank fails one of the five ratios, its balance sheet would automatically be subject to an Asset Cap (similar to the one that Wells Fargo has been subject to since early 2018). But, such Asset Cap would be automatically lifted the minute the bank would again meet all five ratios (just like a trader responding to a margin call and posting more collateral).
The above procedure would have prevented SVB to grow totally out of control, loading on high interest rate risk long-term securities, drying out its liquidity, and going bust in a hurry. It could also prevent hundreds of other potential SVBs from blowing up instead of keeping relying on sheer luck.