WRESTLING WITH HIGHER LOAN LOSS RESERVES

In a presentation at a state banking conference this week, we were publicly questioned on our data for loan delinquencies. The question that arose is that our delinquency data is higher in many cases than actual loss history at many banks. For example, we show multifamily housing loans with a forward delinquency rate (analogous to our probability of default in our loan and credit model) of approximately 1.24%. The banker who asked the question pointed out that this is above the historical average that banks show on their call reports. At face value this is true. However, the reconciliation of this number sheds light on the loan loss reserve calculation. Off the Call Report, the average 30+ day multifamily delinquency is approximately 71bp. This number takes into account a whole range of assumptions that need to be factored in when substantiating loan loss allowance. When BIG puts forward a 1.24% number, this is a static pool calculation. That is to say, this number is devoid of any loan growth and prepayments. These 2 large differences are considerable when banks are growing at a 20% clip, yet are only reserving loans in the 1% range. When examined on a static basis, the effective loan loss allowance comes out to less than 70bp. With every new loan that is placed on the books, the historical loan loss number looks smaller and smaller. Conversely, when a multifamily project is running into problems, refinancing or equity infusion is an option for the owner. Here, in roughly 15% of bank multifamily delinquency cases, banks were successful in getting more equity into the project and either refinancing the debt (within the institution), or driving the customer to another institution. This smart workout move means that historical delinquencies (as a percentage of loans) is underestimated in stable markets. In stressed markets, particularly ones with higher interest rates such as in 2000, refinancing and equity opportunities dry up and prepayments dramatically slow down. In addition to adjusting for growth and prepayments when calculating loan loss reserves, the default rates that BIG utilizes in our loan pricing and credit stress model are forward looking measures based on correlations to interest rates over the last 20Ys. Looking at correlations and then using that data to project forward, based on economic data (such as interest rates, production, inflation, etc.), creates a completely different picture than looking at historical rates over the last 5Ys. Loan losses have been at record lows due to a strong economy and falling interest rates. Therefore, using unadjusted historical data underestimates reserves. One methodology for banks to explain higher loan loss reserves to their accountants, can be accomplished by looking at a static pool, adjusted for growth and prepayments and using historical correlations (projected forward).