C. Myblog

Interest Rate Risk Prepayment Assumptions Vs. Observed Behavior

March 7, 2013
When you or your examiner are evaluating assumptions used in your A/LM model to quantify interest rate risk (IRR), is the objective to prove the assumptions are right?  Or, is it to prove that the assumptions are reasonable and (when possible) supportable?  According to NCUA’s IRR Questionnaire, the answer is you and the examiner “should ensure the assumptions are reasonable and well supported.”
There are a number of assumptions that go into modeling that can be based on observed behavior.  One example is that a credit union can calculate the loan prepayment speeds it has actually experienced recently.  However, simulating what may occur in the future is not an exact science.  Rather, it is more the art of considering past member behavior, recognizing the uncertainty of the future and acknowledging that how members have responded in the past may not be an accurate predictor of how they will act in the future.
Additionally, how prepayment speeds might change in a rising rate environment cannot be observed.  Recent history provides no information as to how 3.5% first mortgages might behave if rates are 300 basis points higher.  The last time long-term rates increased materially (2004 to 2006) LTVs were increasing and the consumers’ balance sheet was very different than it is today.  It may not be worth the incremental benefit to study this time period.
Since no one can accurately, consistently forecast human behavior we believe a prudent IRR management process includes, minimally, annual stress testing of key assumptions.  For example, how could your earnings, risk to earnings, net interest income or NEV change if prepayment speeds are materially faster or slower than actual historical or industry benchmark data suggests?  Understanding how sensitive your results are to changes in key assumptions can prevent your credit union from being blindsided if the future turns out differently than expected.

Start typing and press Enter to search