Our banking industry has our loan loss allowance provisioning almost exactly wrong. To quote our favorite show, Game of Thrones, winter is coming. We don’t know when the cold weather will be here, maybe the first part of October or maybe around Halloween, but we know it’s coming. Since winter is coming, the question arises - How many coats should we have for the winter? For that answer, we simply look at the data and see what the average daytime temperature over the past three months has been, and we can see it is a warm 81 degrees. Based on the data, we can then conclude that we need zero coats for the winter. We should be fine come January.
Winter Is Always Cold
Of course, any banker can see the mistakes in our coat methodology, yet that is roughly how our industry arrives at our allowance for loan and lease loss (ALLL) reserve. Many banks improve on the methodology by looking back longer and taking into account last winter, but the data still gets dampened by the good / warm times. Right now, banks are still reducing their loan loss provisions as the economy is relatively strong and loans are performing. That is a problem as risk is increasing not decreasing. In some markets, commercial and residential property prices are well above their peaks of 2006. Let’s consider markets like New York, Miami, San Francisco and Seattle. In these markets, prices can be 20% to 40% above their peaks of 2007. Can this continue? Maybe, just like we can have two more months of warm weather, but it is not likely to continue for an extended period of time. Cycles happen and when they do, banks are going to want a warmer coat.
Compounding the Problem
Part of the issue is how banks handle risk management. It is bad enough that property appraised values are at a near-record level in many markets, yet loan loss provisions are now back towards their lows, similar to how they were in 2007 before the downturn (below).
To compound the problem, as of August, loan pricing in some markets are back to their tightest spreads, comparable to 2007. Pricing of Libor + 1.50% is becoming more common. Underwriting has loosened and concentrations to commercial real estate are increasing. Commercial real estate loans are now 75% of total risk-based capital and construction lending is back to 16% of total risk capital. Finally, interest rate risk has never been greater, as banks now have the longest duration in the history of banking once you take into account index, floors, caps and prepayment penalties. All this adds up to increasing risk at the loan level and at the balance sheet level. This is at a time when margins continue to contract and are set to further contract even if rates rise in the short and intermediate term.
While banks are highly correlated to real estate prices, let’s just isolate commercial real estate prices for the sake of simplicity. A regression of more than 900 banks in a study by the Bank of International Settlements from 2005 looks at the impact of commercial real estate on bank performance. Their coefficients have some very practical applications for modeling. For example, as can be seen below, the correlations of commercial real estate prices have a larger impact than macro-economic factors such as GDP. For every one-standard deviation of expansion (about 10.85% growth) in real asset prices, bank lending increased 1.74%, return on assets increased 0.10%, margins were reduced by 0.10%, non-performing assets decreased by 0.22% and ALLL dropped by 0.05%. This is greater than the impact of 1 standard deviation of production movement.
The implications for banks are that in good times, banks reap positive performance, however in doing so; they also set themselves up for hard times because the opposite is true. In times of falling prices, the above correlations also hold and banks experience shrinking assets, lower earnings, more asset problems and higher ALLL. What happens if we reverse these correlations?
A Better Model
The industry adopting the new Financial Accounting Standards Board (FASB) Current Expected Credit Loss (CECL) Model is a step in the right direction since it forces banks to be more forward looking. To see if this is true, we built a rudimentary model based on the above correlations and tied ALLL to property prices (and as a result, property level net operating income). Thus, every time net operating income was expected to go up, we increased the reserve, while when net operating income was expected to drop, we decreased the reserve.
What we got was a model that back tested as follows:
As can be seen, our model not only moderates some swings caused by overestimation, but it also moves largely in reverse of how most banks handle their ALLL now. Interestingly, the model also allows us to optimize reserves and ask the question: how much should we be reserving given our current loan mix and historic underwriting exposure? The output was that while banks reserved an average of 1.98% of their net loans, they should have reserved a lower amount to the tune of 1.69% (below). Interestingly, most of the issue came in 2009 and 2010 when banks were largely using the worst case from the downturn to fuel their model.
Does The Model Work?
As a test, we reran cash flows for the industry to see if an ALLL model largely based on the inverse of today’s methodology could reduce risk. We then compared them to actual results to produce the below graphic:
The inverse ALLL Model does serve to moderate income and reduce earnings volatility by 13% which can graphically be seen above. This is a step in the right direction and places banks in a more proactive and forward looking position. This model could be refined even further to include forecasted data that has a better fit (such as effective rents) and be expanded to include other lending lines such as residential real estate (a huge missing piece).
Given the empirical data, we believe banks that face major metro markets should be increasing reserves, not decreasing them. We think the past extraordinary period of low rates has inflated a real estate bubble in many areas and banks are at risk. Current loan loss reserves have been dropping and instead of a 1.38%, banks should be closer to 1.94% and rising. We all know winter is coming and while we don’t know the severity, we know it will be cold. Just like basing your wardrobe on the past will leave you out in the cold, so will having your historic data biased by some of the best credit times in banking. Incorporate more forward-looking measures in your loan loss allowance methodology and you will find that you will stay warmer during the winter months.
CenterState Bank is a $4B community bank in Florida experimenting their way on a journey to be a $10B top performing institution. CenterState has one of the largest correspondent bank networks in the banking industry and makes its data, policies, vendor analysis, products and thoughts available to any institution that wants to take the journey with us.