Richard Holden: How being 'cool' made SVB a bad bank, and why we need better data and tools
Luohan fellow Richard Holden discusses inherent problems in SVB's business model, what this unfolding crisis teaches us about risk and moral hazard, and how better use of data and AI tools may help
Luohan Academy fellow Richard Holden is the President of the Academy of Social Sciences of Australia and a Professor of Economics at UNSW. He is a much sought expert on economic and financial issues and recently wrote articles on SVB in the Australian Financial Review. I caught up with Richard to speak on the economics of the event after his first op-ed How trying to be cool killed off Silicon Valley Bank which also formed the basis for an FT article SVB shows why we should worry about a ‘cool’ bank.
The bank where everyone knows your name
Silicon Valley Bank was a very unique bank catering to the similarly unique and rarefied world of technology firms, their investors, and employees. It cultivated deep relationships in this niche, and by all measures was very good at gaining deposits from them. Stories abound about SVB bankers showing up with sparkling wine at house closings, lending out event space to their start-up customers, or offering cut-rate concierge personal loans to founders.
As the tech industry boomed over the last decade, so did SVB. It deposits and assets rose almost 10-fold in a decade, with the fastest growth during the pandemic tech industry boom. For an industry steeped in an allegedly unique culture and ethos, a bank that was “our bank” probably appealed well beyond mere financial considerations. To Holden, this lead to a “herd mentality among participants in the tech firms in the Valley, where it somehow seemed to be cool to put your money with SVB, rather than some New York bank.”
When success begets failure
Holden strongly believes SVB’s success catering to the technology industry was not a feature, but rather a huge bug. Much has been said about what the Diamond-Dybvig model1 tells us about the “self-fulfilling prophecy” psychology of bank runs. But there’s more insights from economics. For one thing, economists have shown that banks fulfill a crucial matchmaking role due to their information advantage. Bankers and their relationships should not only win deposits, but find profitable lending opportunities. Indeed, research by Diamond’s University of Chicago colleague and former RBI governor Raghuram Rajan showed that despite decades of online banking, around 80% of loans under $1 million still went to borrowers less than 50 miles from the bank’s branches2. SVB did not seem to have an advantage here in lending. As of the latest financial reports, over half of SVB’s assets were in fixed-income securities such as Treasuries and Agency MBS that any of us can buy in secondary markets. No special banking skills required. Even the loan book was over 50% in relatively short-term loans for VC capital call lending, a highly niche business.
Obviously it is understandable why a bank serving a speculative and highly cyclical industry may not be keen to lend long to that industry. This is exactly why tech start-ups have historically been funded with the riskiest of investor money: venture capital. So I see why SVB stashed its massive deposit inflows in “safe” securities - defined as having little credit risk.
But of course, this strategy had interest rate risk that should have been understandable to the most junior banker. Because SVB catered to an industry that was tied to the rate cycle, both sides of its balance sheet had the same risk. This made SVB a particularly uneven ship, that sank when a particularly big wave hit. Depositors were no longer growing balances as fundraising slowed, and fixed-income securities dived in price as interest rates rose in the fastest pace in recent history. Worse yet, a large percentage - over 90% - of deposits were uninsured deposits above the $250,000 limit. Economic theory tells us that these deposits are the ones prone to bank runs. This situation was similar at two other banks recently resolved by authorities - Silvergate and Signature, both of which served the highly volatile crypto industry.
Blocking the dominos
After the run began, authorities devised a response. They used FDIC powers to guarantee all deposits at SVB and Signature Bank, and rolled out a Fed facility that lent against many of those fallen fixed-income securities at par.
So what are some initial assessments of the current situation? Holden thinks the authorities have made justifiably reasonable decisions, albeit in a bad situation. To try to get back to the Diamond-Dybvig non-run equilibrium, they need to show the market enough financial power to mean depositors will not lose their money. Although investors in the banks were largely wiped out, even protecting uninsured depositors from bad decision creates moral hazard for the future. But Holden thinks the trade-off might be worth it. To prevent contagion, “we got to live with some amount of moral hazard in the banking system.” It may be the classic second best optimal solution that is the winner.
Economics is about trade-offs, so what are the potential trade-off? One fear is that this crisis and the subsequent government action might lead to a natural consolidation. Banking is about collective confidence. If the “too big to fail” banks are seen as guaranteed, they may attract more assets and naturally swallow up competitors who are now less viable without a government imprimatur. This potential scenario would not be alien to Holden, whose native Australia has four dominant banks. Ceteris paribus, Holden says of this dynamic, depositors don’t do as well and shareholders earn a better risk adjusted return. Societies need to think about the balance and trade-off. As Holden puts it: “Do we want to have a more consolidated and potentially, at least with good stress tests, potentially safer banking system, but a less competitive one?” In theory, we might be able to give customers the option of higher yielding but riskier banks, without an explicit guarantee. But in practice, this crisis showed how difficult that commitment is when faced with contagion.
How technology can potentially help
After covering these problem that got us here, Holden and I discussed the possible role of technology in helping. In particular, Holden believes that data combined with AI technologies might be helpful in making stress tests better and banks more transparent: “Data combined with some of the modern machine learning methods that are used by economists and computer scientists and other social scientists, may be incredibly helpful in some of these stress tests [given] that the Fed in principle has access to all the data that a bank has.” To me, it also seem a simple case for more technology in risk control. We have a situation where regulator and risk manager are often overwhelmed. The nature of how we model risk is also overly simplistic. People have blamed lobbying by SVB and other banks to avoid stress tests. Yet, Douglas Diamond of the Diamond-Dybvig model has recently noted the stress tests used only tested for 10 year benchmark interest rates rising to 2%. SVB - with a capital level higher than JP Morgan, would have passed those tests.
Holden believes stress tests need to be more rigorous and broader, testing many multifaceted scenarios rather than a rote checklist. While there is no substitute for judgement, being more rigorous is something that technology can help with. If regulators are always overwhelmed and understaffed, AI is a natural aid. Beyond regulators, I also think about customers themselves. The prevailing narrative right now is that you cannot expect an average individual or even business treasurer to have the time or knowledge to monitor their bank’s safety. Perhaps smarter AI, or even adversarial AI, will allow average customers to be more savvy.
Of course, the lesson from all crises is that we should never be over-reliant on any fixed way of thinking. Human judgment that evolves, and ultimately political will, will always be needed to head off crisis.
Holden also keeps an open mind about technologies such as blockchain and digital currencies. They may have a role, although it’s too early to tell how they would factor into a crisis like SVB without knowing how we deploy them. He has an upcoming book on digital currencies. That will be a conversation for another day and we look forward to welcoming him back at Luohan Academy.
Diamond, Douglas W & Dybvig, Philip H, 1983. "Bank Runs, Deposit Insurance, and Liquidity," Journal of Political Economy, University of Chicago Press, vol. 91(3), pages 401-419, June.
JOÃO GRANJA & CHRISTIAN LEUZ & RAGHURAM G. RAJAN, 2022. "Going the Extra Mile: Distant Lending and Credit Cycles," The Journal of Finance, vol 77(2), pages 1259-1324.
 | A guest post by
|