17th March 2023
The basic facts surrounding the collapse last week of Silicon Valley Bank seem clear. During 2021, the bank’s deposits increased by $89bn and the majority of this money was invested not in customer loans but in fixed-rate bonds, mainly mortgage-backed securities.
At the end of 2021, these bonds accounted for $127bn of the bank’s total of total assets of $211bn. The rise in interest rates during 2022 obviously reduced the market value of these bonds, meaning that when the SVB began to suffer a material outflow of deposits, repayment would only have been possible by realising what might otherwise have been passed off as only a temporary and “theoretical” loss – i.e. something that would have reversed over time had the bank been able to hold the bonds (and, of course the deposits) until their final maturity.
Without a shadow of doubt, SVB’s collapse constitutes a catastrophic failure of ALM risk management, so it is pertinent to consider how on earth it could have happened and why such a blatant balance sheet risk was tolerated not only by the risk management function but also by the whole of the bank’s senior management team. Did they not understand the risks or did they just decide to ignore them in the belief that somehow they would not crystallise?
To be clear, investing assets at a fixed rate for a given period of time – in this case over 10 years – exposes you to rates rising unless fixed-rate funding of equivalent maturity can be raised. Or at least manufactured by interest-rate swaps – i.e. swapping fixed cashflows for floating. This is pretty fundamental.
So what were people thinking? Until investigations have been completed and published, no one can really know. One possibility, however, is that they naively believed – or were persuaded to believe – that because the bulk of the deposits were non-interest-bearing ($126bn out of $189bn at end-2021), they already constituted the necessary fixed and permanent funding of the $127bn of fixed-rate bonds. In other words, the argument may have been that any rate rise would not matter much from an income perspective because they paid nothing on the deposits anyway, while a potential fall in market value of assets was equally irrelevant as there was no intention to sell them. They may even have thought that investing fixed was a protection from rates falling.
Now, so called “structural” or “margin compression” hedging of non-dated deposits (NMDs) has been fairly common practice among larger retail and commercial banks for years. The big difference, however, is it is only ever appropriate if you can be absolutely sure that the deposits will remain with you for at least the term of the fixed-rate assets and that they will not need to be re-priced if rates rise. To ensure these conditions are met, normal ALM practice, requires the following:
Whether any of these points were ever considered is unknown at this stage, but if the risk function did allow itself to be persuaded that this was an appropriate structural hedge – and this is just a hypothesis – then, far from it being an excuse, it arguably shows it in an even poorer light. Being either unaware of a risk or being powerless to challenge it are serious enough failings but endorsing, implicitly or otherwise, a simplistic risk management technique without undertaking even the most basic checks is almost worse. As Alexander Pope pointed out over 300 years ago: “a little knowledge is a dangerous thing”.
Another point worthy of note is that SVB in its 2021 annual report did publish the results of its internal Economic Value of Equity (EVE) measurements and, indicated, in the commentary, that it had in March 2021 entered into some swaps to reduce it – presumably paying fixed. This at least shows that there was some awareness both of the risk itself and an acknowledgement that it was getting too big. This, however, raises more questions than it answers:
Interestingly, but probably not surprisingly, SVB did not elect to publish its EVE sensitivities in the 2022 annual report and accounts – it merely presented some 12-month NII sensitivities.
All this, however, is hypothesis based on the few facts and number that are available. The wider point is that EVE risk in banking books should be far, far lower. Furthermore, whatever appetite might be set, it should be solely for the purposes of accommodating peak customer flow and associated operational leads and lags – it should not be there not to facilitate open risk taking. The job of the risk function in a banking book is not just to check compliance with any formal limit but also to ensure any risk positions – even within that limit – have been incurred for good reason and will be closed shortly. To re-iterate, this position was inherently speculative – in order to achieve a slightly higher coupon by investing in fixed rather than floating-rate bonds they were essentially betting that over $100bn of non- interest bearing deposits could be retained regardless of any external events (e.g. a rise in rates), or, alternatively that rates would not rise.
So much then for what may or may not have happened at the outset. What then transpired was, of course, that the non-interest-bearing deposits started to be withdrawn. The fixed asset position consequently became increasingly funded by variable as opposed to no--interest-bearing deposits thus crystallising the inherent risk. It is relevant, therefore, to consider what the drivers may have been to this outflow of funds.
First, the bank obviously suffered a run when its depositors finally woke up to the fact –or were told by their advisors – that there was something seriously wrong. This, however, is not particularly unusual. Banks can only operate and survive if they continue to enjoy the confidence of their depositors but this can evaporate in hours. The driver of deposit withdrawal in recent weeks was simply fear and this was entirely understandable as, at the time, the maximum protection any depositor could have hoped for was $250,000; most deposits were clearly much larger.
Perhaps more instructive, in respect of lessons to learn is to think about what might have been the drivers of the more gradual loss of non-interest-bearing deposits that occurred beforehand when, presumably, most depositors still imagined that the bank was solvent.
Comparing SVB’s end-2021 and end-2022 balance sheets, it can be seen that total customer deposits fell from $189bn to $173bn. Even more significantly, the mix altered: non-interest-bearing deposits fell from $125bn to just $81bn while interest-bearing deposits increased from $63bn to $92bn. What, then, prompted this outflow during 2022 of over a third of SVB’s non-interest-bearing deposits?
One obvious driver is that depositors were simply proving to be rather more rate sensitive than had previously been imagined and many were probably moving their funds into interest-bearing accounts either at SVB or at other banks. As discussed above, large deposits whether from high net worth individuals or from companies with large temporary cash surpluses are more likely to be rate sensitive, so assuming that they will all be retained in the event of rate rises is foolish and highly dangerous.
Another possible driver was what we might term sector concentration risk. From the little we know, many of the deposits came from the “tech sector” - comprising both operational deposits from the tech companies themselves and, by all accounts, individual deposits from their owners and other stakeholders in that industry. Some of the decrease in total balances may, therefore, have been the result not of conscious decisions by deposit holders to move their money elsewhere to get a better rate, but simply reflective of the fact that, collectively, these companies and individuals had rather less ready cash by the end of 2022 than they did at the beginning.
Again, the lesson here is to understand the deposit base. For a large retail commercial bank with a well-diversified depositor base, a decrease in deposits from one sector is likely to be largely offset by an increase in other deposits – the money has to go somewhere – but a bank focused on serving just one sector of the economy is going to find its fortunes rise and fall in line with that sector. It cannot, therefore, make the same assumptions about aggregate deposit stability as a more diversified bank can. This is probably a factor that, to-date, has received relatively less attention in the management of IRRBB. While practitioners and regulators rightly focus on the risk of deposits moving or re-pricing in direct response to an interest-rate change, it is also important to consider other potential drivers of deposit loss. Following the collapse of SVB, it would seem that sector concentration risk on the liability side is probably something that needs to move up the IRRBB agenda.
What lessons then can be learned?
Share "SVB: A catastrophic failure demystified" on