Why doesn't the BancPath Model use Actual Prepayments on Loans as the standard methodology?

It is often assumed that actual bank data is always better than “modeled” data. While we agree that this is true for most modeling assumptions, it is not necessarily the case for the Prepayment Assumption used for Interest Rate Risk. The reasons are as follows:

•  Actual Prepayments infer a rate scenario based on the actual data (and the rates movements that have occurred as the actual data was collected) which may or may not apply to a “base case” scenario.
•  Actual Prepayments are often overstated when LOC and other non-traditional loans are present in the data set, as these instruments can (and often are) funded and paid off several times in the course of their contractual term. 
• It is difficult to interpret historical data for changing rate scenarios in the future. For example, what would be the factor for converting historical prepayments to a +/- 100 bp’ scenario? +/-200? Etc. Historical “actual” data does not translate easily to future rate scenarios.
• Models have already been developed and tested in the MBS markets to track prepayments in both rising and declining rate environments. And while these are not perfect (we use a “factor” to adjust these prepayments based on the loan type) they are considerably more reliable than actual prepayment history tends to be.

For these reasons, we have elected to keep the “standard” prepayment methodology using the prepayment models currently being used

Tags: Loan Prepays, Model Assumptions, Prepayment