Guest Opinion: ALM Model Assumptions Need to Be Tested
An asset-liability management model makes a number of complex calculations for computing your credit union’s interest rate risk. The model’s results depend on the assumptions and whether the model was set up correctly to handle them. An ALM model summarizes results in a variety of standardized formats, many of which allow you to assess the interest rate risk exposure of your credit union. With all the complexities of an ALM model, you want to know that is it incorporating assumptions the way you intended. It’s important to look inside the black box to assess your model’s effectiveness. One way to do this is evaluate the model’s output for reasonableness. A good place to start would be the fair value matrix report.
The fair value matrix report lists detailed categories in the left column followed by columns showing the fair value computation for the flat rate scenario and each rate shock (up or down 100, 200 and 300 basis points). This report, while important, is difficult to analyze. For example, you want to know if the model is incorporating prepayment speeds as intended, and if so, what is the impact on fair value. Ideally, if you can get this report in a spreadsheet format, you can perform a few calculations that will make the results easier to analyze.
The first calculation expresses shocked fair value results in a format similar to that used for bond pricing. For example, a bond trading at par would have a price of 100 while one trading at a discount may have a price of 98, and so on. Compute the ratio of flat to shocked fair values for each category under each scenario and multiply by 100. Let’s use a hypothetical portfolio of fixed-rate 30-year mortgages. If the fair value of the first mortgage portfolio under flat rates was $1,200,000, and its fair value in a plus-100 basis point rate shock was $1,147,200, the calculation would be (1,147,200/1,200,000) * 100 = 95.60. The results are in the Table 1.
By using a common base, Table 1 allows you to assess interest rate sensitivity for each category at a glance. You can also get an idea if the model’s results make sense. For example, you would expect fixed-rate loans to decline in value as rates increase and the intensity of decline to vary based on the underlying characteristics of the portfolio such as loan maturities, payment frequencies, interest rates and prepayment assumptions.
It would also be beneficial to express the impact of rate shocks in terms of its percentage change in fair value for each rate shock category. To do this, subtract 100 from values computed in the table. For example, the +100 column for first-mortgage loans in Table 1 would be computed as 95.60 – 100 = -4.40. So the fair value of this category of loans would be expected to decline 4.40% if rates were shocked up 100 basis points. The results are presented in Table 2.
The format in Table 2 makes it easier to analyze the effect on fair value of different rate shocks for each category. Finally, perform calculations to express the results in Table 2 as the total percentage change per 100 basis points. The values for +100 and -100 columns in Table 2 remain the same. The value of the +200 column is computed as -10.10/2 = -5.05 and the value of the +300 column is computed as -16.17/3 = -5.39. See Table 3.
Table 3 tells more of the story than the previous two tables. What first stands out is that rate sensitivities grow as market rates increase. In this case, growing sensitivities are due to declining prepayment speeds as rates rise. Slower prepayment speeds mean it will take longer to receive the full loan principal, which results in a lower fair value. The results thus indicate that the model is factoring in changing loan prepayment speeds as rates increase or decrease. It doesn’t tell us that the changes in prepayment speeds are reasonable, just that the effects of higher or lower rates on prepayment speeds are considered in the calculations.
Evaluating rate sensitivities per 100 basis points is an ideal way to determine if prepayment speeds and decay rates are being incorporated in your ALM model as intended and to assess the reasonableness of the results and overall effectiveness of the model. It allows for a more apples-to-apples comparison between rate sensitivities and between categories. More importantly, it allows you to zero in on possible errors that may exist in the model’s set-up, assumptions or data inputs. By looking inside the black box you can gain a better understanding of the interest rate risk at your credit union and uncover possible errors before the results are summarized and evaluated against board-approved limits.
Brad Smith is president of ALM Risk Management Consulting LLC.
Contact 317-200-3150 or firstname.lastname@example.org