An alternative take on why Silicon Valley Bank failed
As in 1994 – the last time interest rates were raised aggressively – one asset in particular has triggered turmoil for banks.
To start in the world of fixed income in late 1994 was to arrive at the scene of a crime where all that remained were chalk outlines and grim faces. The Fed had hiked rates aggressively, surprising the markets, and the resulting stresses caused significant turmoil in the financial sector.
One of the main culprits in the volatility turned out to be mortgage-backed securities (MBS). As rates and yields rose, US homeowners, who mostly had fixed mortgage rates, stopped prepaying their mortgages, since they now had a good deal and did not need to pay down a low-interest rate debt.
The result was that the duration – roughly described as the weighted average arrival time of cash flows from a bond – extended massively. Unfortunately, duration also describes the sensitivity of a bond price to changes in interest rates. As Treasury yields rose, in response to the Fed hikes, MBS durations soared and the resulting losses were significant.
Large financial institutions, which did not want to have increasing duration in a bear market, hedged their positions by shorting Treasuries and Treasury futures, further pushing up yields and causing a positive reinforcement loop.1 This feature of MBS is known as “negative convexity”. It cannot be effectively hedged with traditional asset liability-matching tools like swaps and futures, but needs non-linear convex instruments like options and swaptions, which weren’t widely in use at the time.
By the close of 1994, the Fed was heading for a soft landing and the market forgot the tumult as it settled back into a prolonged 30-year bull market. I spent the next few years modelling yield curves and looking at MBS hedging effects, but it was akin to looking for ghosts. Eventually the scene faded from memory as credit crises, emerging market defaults, LTCM and the dot-com bubble burst upon the markets.
2023: the crime returns
The scene cuts to 2023 and a new chalk outline has appeared, signifying the demise of Silicon Valley Bank. There has been much speculation about the ultimate causes of this surprise event. Most forums have clearly laid the blame on a surprising lack of basic asset-liability management. The bank may have been long duration in a rising rate environment, resulting in losses that spooked depositors. But, for those who have managed to remain in this business all this time, the crime scene is eerily reminiscent of 1994.
For SVB, we will take it for granted that the source of difficulty was from the portion of the balance sheet that was designated as securities “held to maturity”. The mark-to-market valuation changes of these securities are not reported in the profit and loss of the bank, and they are kept on the balance sheet at book value. A quick look at SVP’s financial statements, specifically to the securities held in this category at the end of 2022, reveals the following2:
The largest portion of this portfolio is held in fixed-rate agency MBS pass-through securities3, the same murder weapon from the 1994 crisis. The total duration of this portfolio was reported at 5.7 years, which indicates that these are largely fixed-rate, since floating rate would have durations below one year.
A quick look at the ICE BoA MBS index effective duration illustrates the point. The rise in duration from just under two years to over six years constitutes the largest rise in duration in the history of the MBS market4. This is of course accompanied with one of the worst bear markets in recent memory for Treasuries. Putting the two together is the perfect storm and this is what has likely scuttled the SVB balance sheet.
Unlike Treasuries, MBS durations and interest rate sensitivities move around a lot over time
Using fixed rate MBS holdings (held-to-maturity) data from the SVB financial statements and overlaying MBS index durations and 10-year Treasury yields gives an interesting hypothesis of the crime scene.
The massive accumulation in fixed rate MBS occurred before most of the duration extension happened. As yields continued to rise in 2022, the interest rate sensitivity of the holdings had already increased significantly. If the index duration is a good approximation, the riskiness of these holdings effectively tripled between the beginning of the accumulation and the eventual collapse.
SVB built holdings of mortgage-backed securities just before rates and duration started to climb
While it is far too soon to present the ultimate motives and modus operandi to a jury, it is important to study the clues available in order to get a sense of when the killer might strike again. If the MBS are indeed the culprit, using basic interest-rate hedging tools like interest-rate swaps would only partly have offset the prevailing risks lurking in the balance sheet.
Understanding the contributing factors to the SVB demise is important if we are to understand any potential contagion. Small banks or "woke" hiring policies may not be the villains we are looking for.
1 For a description of this period, please see “Mortgage-backed mayhem: Naive investors, aggressive dealers, baffling instruments—the market was a disaster waiting to happen. Will it be next time around?”, Michael Carroll and Alyssa A. Lappen, Institutional Investor July 1994, pp. 81-96
2Page 66 of SVB Financial Statements for December 31st, 2022
3There are some assumptions here that are important to note. We have assumed the Agency-issued MBS and the fixed-rate collateralized mortgage obligations are the main instruments with negative convexity.
4We don’t know the composition of the fixed-rate MBS on the SVB balance sheet, so this index duration should only be considered indicative as an approximation. Note the effective duration is only available post 1995 and so modified duration is shown prior to this period.