Pub. 7 2016 Issue 1

www.wvbankers.org 8 West Virginia Banker Back Testing and Validation of A/L Reporting Systems By Jeff Caughron and Chris Wilson, The Baker Group LP F or a number of years, bank regulators have focused on the adequacy and soundness of ALCO processes used by banks to manage interest rate risk exposures. With respect to the reporting system or model that is used for A/L analysis, there are certain key points that are generally considered. These will often include the following:  Does the model possess sophistication and complexity that is appropriate for the institution?  Are assumptions back tested regularly?  Is data-input accurate and current?  Are model assumptions reasonable and updated?  Has the model itself been tested and validated by an outside auditor? Two of these, back testing and model validation, are especially noteworthy. These are entirely different processes, but equally important to the overall in- tegrity of any modeling process and the resulting usefulness of the reports. It’s important to understand the distinction between them. The starting point for any good A/L model will be accurate input of key bal- ance sheet information such as balances, yields, reinvestment rates, and the con- tractual re-pricing or maturity schedules. These are the known data entry char- acteristics. From there we must build into the model a variety of assumptions in order to account for those things we cannot know in advance. These include time lags, sensitivities or betas, behavior of non-maturity deposits, and prepay- ment speeds, among others. Asset/ Liability modeling therefore involves the input of certain known characteristics combined with certain unknown be- haviors that must be assumed. Gauging the reasonableness of these behavioral assumptions is important to assessing the overall integrity of the system. Back testing is helpful in this regard. What Is Back Testing? Back testing is the process of reviewing the projections of an A/L model after the fact and comparing those projections against actual performance. The results of a back test help us to answer questions such as: 1. Were our projections different from actual bank performance? 2. If so, by how much? 3. What inputs, assumptions, or dynamics might explain the variance (e.g. time lags, the competitive environment, changes in the local economy, etc.)? The ultimate objective of a back test is to give guidance as to if and how manage- ment might want to adjust current assump- tions within the A/L model in order to more accurately represent the performance trends of the bank. A back test, unlike model validation (the topic of the next section), does not need to be performed by an outside auditor. The back test may be done internally or with support from the