Observations from ALM Model Validations: Prepayment Speeds
This blog is a continuation of our series to share observations from A/LM model validations that we’ve performed.
Prepayment speed assumptions impact earnings, net worth, and NEV within an A/LM simulation. Too often, though, people may simply expect that their assumptions have been input correctly and are operating as intended. Some of the issues to look for are:
- Prepayment speeds stay the same in all rate environments. This should not be the case for mortgages. For other types of loans, if prepayment speeds don’t change as interest rates change, ensure that assumption seems reasonable for the particular account
- A loan or investment account that may be tied to the wrong prepayment speed table. Specific account details are often on a separate page from prepayment tables, so linking the two can identify if there are inconsistencies between the two
- Prepayment speeds may be difficult to locate or interpret within the simulation reports
- Modelers may accidentally flip-flop prepayment speeds. When market rates go up, expect prepayment speeds to go down as borrowers become more reluctant to refinance
- Newly added accounts may be assigned to the wrong prepayment table, or may not be set up with any prepayment speeds (it is always good to pay special attention to new accounts)
- Prepayment speed assumptions in the income simulation are different than those used in the NEV simulation when they should be consistent. These different assumptions occur because some modelers use different systems to produce their income simulation and NEV
- Prepayment assumptions provided/expected are not those in the model
Once you have verified that the intended assumptions are being modeled, you may then also want to understand how the results could change if the assumptions were different. For example, what if prepayment speeds actually experienced were 50% more sensitive than those modeled? Conducting stress tests over a range of different rate environments can demonstrate the sensitivity of the results to the assumptions, which may make you, and your examiner, more comfortable with the assumptions being modeled.
