Far too often communication and knowledge gaps between interrelated functions can leave executive management teams frustrated when unexpected outcomes are realized.
Gap1: Sensitivity Analysis
Every MSR, bond or loan portfolio has unique characteristics and embedded risks. Not only do we need to understand the exposure to key assumptions or risk factors(prepayments, default, concentration, etc.), but consideration should also be given to well-developed, probability weighted scenarios that contemplate significant circumstances.
Larger institutions (>$10B) are very familiar with these disclosures but we recommend that the documented and management reviewed exercise be adopted as best practice across all acquisitions and retained portfolios.
Gap 2: Assumption Back-testing
You've provided management the set of assumptions used to evaluate and execute on the trade. Whether based on internal or independent buy-reps, you need to now ensure that assumptions made in your trade execution are coming to fruition in the actual performance of the collateral.
Typically, acquired portfolios are flagged for valuation and MTM purposes but we highly recommend developing granular processes to isolate the pool and back-test performance against assumptions leveraged in the trade; especially important for any buy-hold strategies.
Gap 3: Interest Rate Spreads
Every model has limitations and our ability to model spreads continue to be a challenge. Most of us have migrated to mean reverting spreads over some determined half-life. However, in modeling mortgage rates we tend to focus on primary and secondary spreads in determining the base mortgage rate against which moneyness and related prepayments for specific collateral, 30Yr FRMs let's say, are calculated. More sophisticated models also include the relevant spread and impact to other tenures, the relationship between 30Yr and 5/1 ARMs, let's say.
Hopefully, you're not caught off guard by the expected increase in 5/1 ARM prepayments even though rates on the product remain relatively flat to last year averages.