Do A/LM Modeling Results Match What’s Intended?
Is your A/LM model producing the intended results? During the course of validating other A/LM models, sometimes we see inputs that appear reasonable on the surface but are not producing the intended results.
When conducting a recent credit union model validation, we were asked why the model showed a loss on 30-year mortgages when the credit union intended to show a market value gain. The account was discounted to the yield curve plus a spread with the intention of having the discount rate equal to current offering rates. Given that the yield on the account was materially higher than the discount rate, a market value gain would have reasonably been expected. Digging further into the details revealed the cause of this issue: slower-than-average assumed prepayment assumptions.
The credit union’s decision to slow down prepayment assumptions pushed more maturing cash flows further out on the yield curve, to a point on the curve where the discount rate was actually higher than the portfolio yield. The cash flows further out on the curve were at a sizable loss, enough of a loss to wipe out all of the gains on the maturing cash flows at the shorter end of the curve. In the end, the credit union made adjustments to deliver a result more in line with its expectations of the portfolio’s value.
While this example represents an unusual situation, the takeaway here is simply to stay on the lookout for unexpected results. Always ensure that your modeling results make sense to you. Understand and verify the accuracy of any unexpected results, or make adjustments as appropriate to correct the problem.
