Different, but the Same: What Bank Failures Can Teach

Things had been thankfully quiet on the bank failure front before March 2023.

Since aggressive lending backfired in epic fashion in 2007, toppling more than 500 American banks over the next half-dozen years, collapses had slowed to a trickle. Pew Research reports an average of fewer than four failures per year from 2015 to 2022, and notably none at all in 2021 and 2022.

Then, over a long weekend in March, the industry witnessed the second- and fourth-largest bank failures (adjusted for inflation) in U.S. history. Santa Clara, California-based Silicon Valley Bank (SVB) cratered on Friday, March 10, after a 48-hour bank run, and New York’s Signature Bank followed suit the following Sunday.

Though both were regional banks, the two firms were very different animals. SVB was predominantly a wholesale bank, catering to technology companies and investment portfolios. Signature’s clientele was largely mid-market real estate buyers and cryptocurrency companies.

SAS’s Donald van Deventer: Don’t neglect lessons learned.

SVB’s collapse, while abrupt in its climax, ignored hazards that had been in play for a year or more. Signature was a domino falling in SVB’s wake, with the Federal Deposit Insurance Corp. (FDIC) springing in to make sure it was the only one. So far.

But for all of their differences, the SVB and Signature failures have more in common with each other – and with historical bank failures as a whole.

Why Banks Fail

As experts strive to forecast how the industry will recover from the current crisis, the only fail-safe prediction one can make about bank failures is this: There will be another. Hopefully not next week, next month, or even next year, but no amount of regulation or supervision will ever prevent them entirely.

Despite that sobering reality, there’s hope. Firms can tap the growing wealth of knowledge about understood causes and symptoms to help steer away from the rocks. And they can lean on the rapidly advancing technology capable of turning those lessons learned into actionable insight in near real time.

While specifics vary, bank collapses are almost always predicated on the collision of inadequate risk management and a sea change in the macroeconomic climate – thus, the propensity for bank failures to happen in waves. In the late 2000s, an appetite for high-risk lending ran headlong into a loosening of regulations and a recession, causing a flood of defaults.

S&Ls, Continental Illinois and Concentration

Of course, the circumstances behind the current disruptions better mirror an earlier banking crisis, when between 1986 and 1995 the industry saw 1,043 U.S. savings and loan institutions shuttered. Through the 1980s and early 1990s, speculative loans in oil and gas and investment in developing countries met tightened monetary policy to combat runaway inflation, also precipitating a recession. When Continental Illinois National Bank & Trust Co. failed in 1984 – at the time, the largest bank failure in U.S. history – more than two-thirds of its deposits were concentrated in accounts over the FDIC insurance limit (then $100,000).

SVB mimicked that concentration, a recipe for vulnerability to a run on deposits. Signature’s crypto business accounts demanded a minimum balance of $250,000, the current FDIC limit. Crypto’s implosion exposed Signature to similar risk.

If history is any lesson, the final straw for SVB was one that should have been foreseen. In fact, the bank’s board, despite a long vacancy in the role of chief risk officer last year, seems to have recognized there was exposure. It called 18 risk management meetings in 2022, compared to seven in 2021. But to what effect?

A perfect inflation storm had been brewing since the advent of the COVID pandemic in 2019, with supply chains fractured and deposits mounting. Many corporations took record profits while amassing massive debt. It wasn’t a matter of if the Federal Reserve Bank would raise interest rates, but when. Yet SVB’s capital stayed parked with long-term fixed rate Treasuries, while its depositors had a world of other options.

There was a huge amount of data from failed banks to learn from. Somehow, SVB and Signature didn’t.

Studying Failure and Default Probabilities

Studying data from bank failures gives valuable insight into their underlying causes. With data becoming ever more all-encompassing, granular and timely, it can also help banks identify symptoms, and data patterns (like deposit outflows), that predict default danger.

In a spreadsheet-based world, this data was overwhelming. As recently as a few years ago, stress testing and scenario analysis capabilities faced the limitations of the number of hours in a day. But recent advances, especially in the velocity of calculation and the self-learning capabilities of artificial intelligence (AI) and machine learning in the cloud, make more exhaustive analysis possible.

It’s especially true of invaluable Monte Carlo simulations, wherein a range of random values for particular variables predicts instances of failure. In order to avoid SVB-level risks, banks must simulate the value of the assets, the value of liabilities, and count the scenarios when the value of assets is less than the value of liabilities. Divide that tally by the overall number of scenarios, and that's the bank’s default probability at that particular time. Inexplicably, many institutions fail to do that critical final calculation.

Whereas the Federal Reserve’s three-scenario risk assessment might be considered sufficient by some, predictive analytics can push that number into the thousands, while AI homes in on failure cases to extract ever more precise and thorough information that can help banks take action to mitigate the risks once hidden in their balance sheets.

Prioritizing Balance Sheet Risk Management

In some ways, the long weekend that marked the end of SVB and Signature Bank seems like a bump on the risk management freeway. But risk professionals with their eyes on the road can see the “Road Work Ahead” signs flashing caution on deposits, with an imperative to shore up asset and liability management (ALM) capabilities.

A lot of institutions wrongly conclude that the regulations represent best practice when, in fact, regulatory requirements are merely a baseline for conscientious risk management. Banks have an obligation to hold themselves to much higher standards. And they must also remember that they work for their shareholders, not the regulators.

So, what are some other steps banks can take to navigate market volatility?

• Invest in advanced risk management tools – and talent. Predictive analytics, AI and machine learning can improve banks’ ability to identify, monitor, score and manage risk. More than just bolstering their technology, banks have a talent gap to fill. They’ll need to lure risk managers to return to interest rate risk roles that waned in the glory years of historically low interest rates. Consider, for example, that the one-factor interest rate models used by many banks underestimate the degree of interest rate volatility between 61% and 83% when compared to a more accurate 10-factor model. That’s a clear indication that the right tools in the right hands can be the difference between preparing for macroeconomic changes on the horizon and reacting to them on the doorstep.
• With renewed focus on ALM, don’t neglect the credit risk lessons learned. Most often, market disruption of this scale starts with credit risk. In SVB’s case, an interest rate shock exacerbated balance sheet management deficiencies, ultimately exposing liquidity risk that led to a run on deposits. Banks must monitor how this impacts the longer-term credit outlook, particularly given the record levels of corporate debt currently outstanding. While credit is difficult to forecast, banks would be wise to implement a transparent model that is regularly updated, ideally daily, with a proven history of being predictive in the past.
• Keep risk management governance, policies and processes sharp. Market conditions and regulatory regimens remain in flux. Simulate, simulate, simulate to a minimum 10-year time horizon. Account for as many adverse scenarios as technologically possible and hone risk management strategies accordingly. A holistic, granular view of cross-spectrum risks, integrated in one cohesive platform, is the gold standard in the face of today’s increasingly precarious risk management climate. Daily balance sheet management is possible with the right tools and team in place.

I published a blog post in 2009, 4 Questions that Every Bank CEO and Board Should Be Able to Answer. Question No. 1 was: “What is the probability of default of our bank monthly for the next 10 years?”

Many things have changed since I wrote that post, but that advice has stood the test of time. Financial institutions can’t rely on luck to survive the current banking crisis. Now, as in the past, they’ll need preparation and diligence, propelled by the most comprehensive and agile risk management strategies, technologies and talent available.

Donald van Deventer is Managing Director of Risk Research and Quantitative Solutions at SAS. He founded and was CEO of Kamakura Corp., a specialized risk data, software and consulting firm which SAS acquired in June 2022. The views expressed in this article are his own.