Close
Bank forecasting has to advance beyond a simple regulatory compliance “tick-the-box” exercise. It must be used as a tool for making strategic decisions.
Customers deposit money with traditional financial institutions, which are then used to fund loans. Yet, they lend out considerably more than banks have available at any given moment, a practice known as fractional banking. On the one hand, a bank’s profitability is based on its net interest margin, which is the difference between interest earned on loans and interest paid to depositors. The bank’s ability to withstand external shocks is determined by its equity, which is the difference between its assets and liabilities.
SVB owned $212 billion in assets compared to around $200 billion in liabilities before to the most recent run on the bank, making it both a successful and secure financial institution. Hence, their equity buffer was $12 billion, or 5.6% of their assets. While it is less than half of the 11.4% average for banks, that is still not too awful.
The issue is that the value of long-term debt, to which SVB was significantly exposed through its mortgage-backed securities (approximately $82 billion), has recently decreased as a result of recent moves by the US Federal Reserve. Several worries were raised when SVB informed its shareholders in December that it had $15 billion in unrealized losses, eliminating the bank’s equity buffer.
SVB reported on March 8 that it had sold $21 billion in liquid assets at a loss and that it would seek cash to make up for the loss. Nevertheless, the news that it needed to raise more capital and even thought about selling the bank greatly alarmed investors, who sought to withdraw almost $42 billion from the institution. Of course, SVB lacked adequate money, and on March 17 the Federal Deposit Insurance Corporation seized control.
There is a lot of information in the macro-finance literature concerning these circumstances, but a decent summary is to expect extremely non-linear dynamics, which means that output can be significantly affected by modest changes in inputs (such as the equity-to-asset ratio) (liquidity). Bank robberies could occur more often and have a significant impact on overall economic activity during recessions.
SVB is not the only bank, to be sure, that is more exposed and at risk to macroeconomic factors like interest rates and consumer demand, but it was simply the tip of the iceberg that made headlines during the last week. And we’ve seen this before, most recently with the demise of Washington Mutual during the financial crisis of 2007–2008. The Dodd-Frank Act, which increased the Federal Reserve’s authority to regulate financial activity and authorized new consumer protection guidelines, including the establishment of the Consumer Financial Protection Bureau, was largely the result of the aftermath, which saw a surge in financial regulation.
Notably, the DFA also implemented the “Volcker Rule,” which prohibits banks from engaging in proprietary trading and other speculative investments. This effectively prevents banks from acting as investment banks and trading stocks, bonds, currencies, and other financial instruments with their own deposits.
The need for science, technology, engineering, and math (STEM) personnel, or “quants,” changed dramatically as financial regulation increased. Because that regulation has an impact on their non-interest expenditures, financial services are particularly vulnerable to regulatory changes, with labor bearing a large portion of the burden. The banks understood that by expanding automation, they could lower compliance costs and boost operational efficiency.
And that’s precisely what happened: Between 2011 and 2017, financial services had a 30% increase in the number of STEM personnel, much of which was due to the increased regulation. Small and mid-sized banks (SMBs) have, on the other hand, found it more difficult to comply with these laws, at least in part because it is expensive to hire staff and develop complex dynamic models that can predict macroeconomic circumstances and balance sheets.
The most up-to-date models for macroeconomic forecasting are outdated econometric models from the 1990s. With the exception of some innovative and experimental methods used, for instance, by the Atlanta Federal Reserve with its GDPNow tool, there is no consensus workhorse model or approach to forecasting future economic conditions, despite the fact that forecasts are frequently modified at the last minute to appear more accurate.
Nevertheless, even these “nowcasting” systems do not include a significant amount of disaggregated data, making the projections less relevant for SMBs that are exposed to certain asset classes or areas and less concerned with the overall status of the national economy.
Forecasting has to transform from a “tick-the-box” regulatory compliance measure to a real strategic decision-making tool. If the nowcasts don’t work as expected, cease making them or find a means to make them helpful. The world is extremely dynamic, therefore we must make use of all the resources at our disposal, from powerful machine learning algorithms to disaggregated data, to assist us comprehend the current situation so that we can act responsibly and avert possible disasters.
Could Silicon Valley Bank have been salvaged with better modeling? Maybe not, but improved modeling would have enhanced transparency and the likelihood that the appropriate queries would result in the appropriate safety measures. Technology is a tool, not a replacement, for effective government.
There has been a lot of pointing fingers and going over the past since Silicon Valley Bank collapsed.
