FDIC Insurance is the holy grail (Right?)

Play audio

In my last article, I discussed the difference between a bank’s performance and probability of failure.  Most bank rating agencies look at performance not potential for failure.  As a consumer or business person, your main goal should be preservation of capital.  With almost 4,000 traditional institutions in the US, you’d be surprised how many banks have over a 50% chance of failure.   In other words, what sort of characteristics are associated with a higher probability of failure.  As a risk officer at a community bank, I set out to see what the probability of failure was for my bank.  I spent months and months looking at dozens of metrics and computing many regression analyses to come up with the right fit. It took me so long to research it, that I realized I couldn’t spend all my time at work preoccupied with this analysis.  I decided to do it in my spare time and look at all institutions not just mine.  From that sprouted up www.bankeroverboard.com even though I had no experience in building websites. 

At first, I used traditional metrics in my analysis, but as I added more and more metrics, I found different variables that actually had a stronger correlation.  Even better still, as banks failed, I realized the variables changed overtime because the reasons changed.  The regulatory agencies have a very systematic tried and true way of analyzing institutions. But what happens if something in the economy, politics, or around the world occurs that has never happened before?  When SVB failed it sent shock waves through the economy, but also through the regulatory agencies.  They had no clue what was about to happen.  In fact, there is no record of formal enforcement action against SVB leading up to the failure.  While most banks fail because of poor loan quality, SVB did not.   A combination of events including a rapidly changing interest rate environment, a group of analogous depositors, known as venture-capitalists, and the rare, but ever-present run-on-a-bank occurred.  Unfortunately for the regulators, history didn’t repeat itself.  I can assure you though, SVB will be a case study for all agencies going forward.  Our analysis takes these factors into account.  In fact, we update our analysis every quarter to make sure the latest issues are accounted for in our numbers.

Truth-of-the-matter is, bank failures are quite predictable (set aside the occasional SVB).  A quick google search or AI search will result in numerous sources confirming this through multiple studies.  So why might depositors still lose their hard-earned money?  In a word….complacency.   If you have FDIC insurance, why should you care about the probability of failure of the bank?  One, you may not understand what accounts are covered by insurance.  Second, you may have never heard of a bank failing because many are bought out, fortunately.  Or third, you think your bank is invincible because you walk in the branch or visit their online account every day.  Lastly and most importantly, doesn’t the US Government bail-out large banks? They have in the past, but that doesn’t mean they will in the future.  Truth is, traditional banks are a safer place to secure your money than most investments, but better to be safe than sorry when risking your hard-earned money. 

Our analysis focuses on traditional banks.  This includes banks of all sizes.  Community Banks and Money Center Banks (sorry, it’s a boomer thing and means large banks). It does not include Savings Banks, Credit Unions, Credit Card Specialty Banks, Bankers Banks, Mutually-owned saving banks, or Denovo Banks.  These institutions exhibit characteristics that are out of the norm and can skew metric results making the probability numbers less reliable.  Credit Unions are particularly worrisome since they do not have the same plethora of financial information available to the public.  Transparency is your friend.  Beware of latest fintech online account promising the world.  It could be a wolf in sheep’s clothing.   

And now to the good stuff.  While our research is proprietary and where our business value comes from, I will share what I can.  We are not statisticians, but have learned enough to pull together a very comprehensive program.  Our value is truly in picking the correct metrics or variables that correlate to bank failure compiled over my almost 35 years of banking experience.  They are not always the traditional metrics you might think.  And they will change overtime as the banking industry changes.  We are constantly upgrading our analysis.  In keeping with proper statistical techniques, our analysis included hundreds of bank failures and, also, banks that didn’t fail over more than 10 years.  The regression analysis is also backed by an AUC analysis.  AUC analysis refers to the Area Under the Curve in a Receiver Operating Characteristic (ROC) curve. It is a metric used to evaluate the performance of a specific classification model. The AUC value ranges from 0 to 1, where:

• 1.0 indicates a perfect model.
• 0.5 suggests the model is no better than random guessing.
• Below 0.5 means the model is performing worse than random.

AUC is particularly useful because it measures how well a model distinguishes between positive and negative classes across all possible classification thresholds. The results of the most recent model are .9941 percent or a strong predictor with a high level of accuracy.  However, I am by no means declaring anywhere near perfection.  Potential overfitting exists, which means a high accuracy on training data but low accuracy on test data and only time will tell if the model continues to get strong.  The predictability has actually improved over the past year as we add new observations.