It is certainly in vogue these days to warn community banks about the risks associated with rising interest rates. Regulators, auditors, and consultants (yes, even me) have all sounded the alarm in one way or another. Interestingly though while many community bankers acknowledge the risks, they are certain that their own bank is adequately prepared. They believe (quite adamantly in some cases) that their bank is asset sensitive, and that the Fed’s extremely accommodative policy is the only thing standing between them and margin pressure relief. I find this very interesting and since we regularly run an A/L model for nearly 200 community banks, I am routinely asking – are you sure?
There are a handful of “big” assumptions that banks use in their interest rate risk models. It might be hard to say which one has the most impact, but I think there is general agreement that the core deposit betas or rate change assumptions can clearly have a big influence on the results.
What assumptions are you using? When running the income simulation a common method of modeling core deposit rate sensitivity is to use a beta factor. Essentially the beta factor defines the percentage of the market rate change that will be applied to the bank’s core deposit rates. It is usually specified as a percentage of the movement of some standard rate like Fed Funds. For example, suppose the bank’s beta factor for interest checking accounts is 25%. If the target Fed Funds rate moved up by +100bp, the model would increase the bank’s rate on interest checking by +25bp. It’s a simple and effective way to model the typical lag between core deposit rates and market rates. The trick is choosing a reasonable beta.
Your bank, like many others, may have chosen a beta that attempts to mimic your core rate behavior in the past. Looking back into historical periods of rising rates you “computed” your beta by comparing the change in your core rates to the change in market rates (perhaps even using some simple regression analysis.) But is that beta factor “right”? Your income simulation measures your exposure by projecting forward your balance and rate behaviors, not by looking back. Does your beta factor reflect how you will behave in the future?
Let me ask the question differently. What if your beta factors are wrong? Lets say for example that currently, using its myriad assumptions - including your chosen beta factors, your model shows your bank is asset sensitive i.e. your margin should improve if rates rise. How much would we have to change the beta factors so that your bank is instead liability sensitive? How sensitive is your bank’s measurement to this modeling parameter? Answering this question is precisely what the 2010 IRR Advisory was referring to:
Proper measurement of IRR also requires sensitivity testing of key assumptions that exert the greatest impact on measurement results.
FIL-2-2010, 2010 Advisory on Interest Rate Risk Management, 01/20/2010, FDIC
Here’s a simple example of a bank that becomes liability sensitive simply by increasing the beta factors used.
In the base case (shown in blue) the bank projects a weighted cost of 0.17% for its core deposits (shown on the “Sub Total” line). Scenario #1 is a +200bp instantaneous parallel shock analysis. The beta factors used are 10%, 10%, and 25% for interest checking, savings, and money market accounts respectively. The projected weighted cost of core deposits is 0.49% in this first scenario. Overall projected margin moves from 4.26% in the base to 4.35%, a positive change which is considered “asset sensitive".
Scenario #2 is the same as Scenario #1 except that we’ve changed the beta factors. Now the beta factors used are 20%, 20%, and 50% for interest checking, savings, and money market accounts respectively. The projected weighted cost of core deposits is 0.81% in this second scenario. Overall projected margin moves from 4.26% in the base down to 4.23%, a negative change which is considered “liability sensitive".
How much is your overall IRR profile driven by your core deposit repricing (beta) assumptions? You should be able do some back-of-the-envelope sensitivity testing like the one I’ve shown above. If you are an A/L BENCHMARKS customer we provide a simple web-based tool that will allow you to run this test yourself. It will make a good discussion topic for your next ALCO meeting.