Author: Joe Rezac, ALM Services Manager, jmrezac@profitstars.com
With football season coming to a close and the seasons changing, my attention has turned to my winter hobby – woodworking. Stepping into the shop after a long hiatus, I’m always amazed by the number of woodworking tools that I’ve accumulated over the years. Despite this, many different tools can often be used for the same job. A hand tool may offer more precision than a power tool, but could take twice the amount of time to use.
Determining reasonable Asset-Liability Management (ALM) model estimates is a lot like woodworking. There should be many tools in your toolbox when you work on your model assumptions, helping you to achieve your goal of coming up with reasonable model results for interest rate risk.
One ALM model assumption that has been in the spotlight recently is the deposit account decay estimate. Until recently, the use of industry-derived decay estimates – like those from the Office of Thrift Supervision (OTS) or the National Economic Research Associates (NERA) – were standard practice in many ALM models and were not overly criticized. Simply using antiquated industry standards are no longer going to cut it. There is no documented rule-of-thumb for what a typical deposit account’s life should be, and there is little regulatory guidance for how to come up with proper decay estimates – other than the method used should reflect the size and composition of your balance sheet. Since non-maturity deposit (NMD) balances play a significant role on the liability side, coming up with a reasonable assumption for NMD life is critical in fair value analysis.
Various approaches can be used to come up with depositor retention based on historical analysis. One popular approach involves looking at a group of accounts that were opened and closed within a particular timeframe and seeing how these account balances change relative to changes in interest rates. But be careful. If you take this approach, don’t stop there. Often this type of analysis results in long retention times. The longer the retention time used for a deposit account, the more of an implied premium exists in fair value calculations, which can skew overall fair value results. Today, long deposit lives are facing strong examiner scrutiny. In many cases, the financial institution simply uses their more rosy study results as a negotiation chip to go back to using the old OTS/NERA decay estimates. When you end up scrapping the results of a third-party study that you paid good money for, that’s an expensive lesson to learn! Equating back to woodworking, that’s like being told you can’t use the new state-of-the art table saw you just bought. Instead, you should revert back to using a hand saw, as it’s less likely to cut off your fingers!
With increased examiner scrutiny a reality no matter which method is used, there are easier-to-understand, less time-consuming, and potentially less-expensive options available. Which options depend in large part on how much historic depositor data is accessible to you. Items such as account open dates, close dates, average balances, and account types can be tracked over time to begin to develop what the depositor behavior is for your institution. Ideally, at least one full business cycle worth of data is needed. If you are limited on historic data, start with whatever is available and track it going forward.
Again, don’t stop there. Your deposit base is dynamic, so the process for tracking its ebbs and flows also needs to be dynamic. Did your institution experience “surge deposits” since the 2008 financial crisis? If so, what percent of each account is represented by this group? Look at data back over 10 years – how many accounts are still open after years one, two, three, etc.? How many are still open today? Now look back at the data over a five-year horizon – what does that tell you? What future events (either internal or macro-economic) will likely change depositor retention from what it was historically? Repeat your process periodically to make sure the information you get from the results stays current. Discuss with your Asset & Liability Committee (ALCO) and examiner to see how the process can be fine-tuned to produce more meaningful estimates.
Once a base set of assumptions is developed, consider estimating how the base assumptions change when market rates change. Traditional statistical analysis has focused on how retention is impacted when market rates go up or down. However, this disregards all of the other macro-economic factors that go into a depositor’s decision to keep money parked at their financial institution or close out the account and do something else with the money.
According to a 2011 J.D. Power & Associates study on U.S. Retail Bank New Accounts, the number one reason people switch banks is life circumstances. Life events such as moving, job loss, birth, death, marriage, and divorce all play a more critical role in a depositor’s behavior than what current market rates are doing. But can this be properly accounted for in a rate-driven model? An argument could possibly be made that the rate of heart attacks increases when market rates drop 300 basis points overnight, but unless you have an actuary on staff, it may be difficult to determine the precise correlation between the two. There is no “easy button” here. How much of a factor rate plays in your depositor retention for each account type is something that your ALCO needs to discuss, document, determine, and refine on a regular basis.
Depending on the size and complexity of your balance sheet, a statistically-based rate analysis may still be part of your procedure for coming up with viable depositor retention estimates, but other approaches should also be looked at as well. At the end of the day, are you comfortable with explaining your methodology to an examiner or a volunteer board member? Whether it’s used for regulatory stress testing or strategic planning, the tool that you are the most familiar with (and is best designed to complete a task) is the one that will most likely achieve the best results for you.
.svg)
