1. In Washington, there always seems to be a discussion about GSE reform, but the likelihood of it actually occurring is anyone’s guess. How do modelers and MBS investors prepare for that type of uncertainty?
Over the past ten years, there have been innumerable proposals to change the functioning of the housing finance system and the mortgage secondary market. As modelers, we want to provide our clients with the tools to evaluate the impact of these proposals on their investments and their business. Our approach has been to keep our model focused on what we can see in the data and address potential changes through other mechanisms.
A good example is the many changes to the refinancing incentive through programs like HARP (Home Affordable Refinance Program). When various proposals were presented and implemented, we produced an analysis to show how various levels of take-up of the HARP program could affect prepayment rates. Once data became available we incorporated the change in behavior into our models.
We believe that this is the best approach for our clients: Make them aware of the potential impact of change, but don’t add speculative features to our models.
2. Over the past year, the issuance of non-agency MBS has begun to increase. Some observers speculate that FHFA Director Watt’s replacement could look to accelerate that trend administratively. With credit risk on the table, have PLS investors been quicker to adopt novel valuation and analytical tools? Do you expect that to change?
While we hope for continued increases in the issuance of PLS, that market is still limited. On the PLS side, the greatest growth in credit risk analysis has been in the evaluation of non-QM (Qualified Mortgages). Given the limited history of these mortgages as a separate asset class and the relatively benign housing market environment, it is difficult to assess how these loans will perform in a stressful environment.
On the other hand, there has been a great deal of activity in the (CRT) Credit Risk Transfer market with securities issued by Fannie Mae and Freddie Mac. The growth in these programs has produced a great deal of interest in relative value and other credit risk measures as traders and investors in the PLS market switched their focus to the CRT market. The recent developments in the CRT market are likely to spill back into the PLS market when issuance increases.
3. How will machine learning and AI transform the modeling in the secondary mortgage market? Are there areas where you’re seeing high-impact use cases today?
As modelers, we always welcome new data and new analytical techniques. For us, much of what is described as machine learning and AI are techniques that build upon approaches that we have used for years. As a result, we see the move toward AI and machine learning as evolutionary rather than revolutionary. We have also seen the limits of technology in the evaluation of mortgages and view solutions that seem “too good to be true” as probably “too good to be true.”
In the mortgage market, the greatest impacts so far have been in expanding the datasets used for credit modeling, including the work being done by VantageScore. We have also seen some inroads in using machine learning to identify potential variables that can be added to mortgage models. To date, the full AI models of mortgage performance have not been able to match the accuracy of human-guided models and fall far short in producing explainable and actionable results.
Our biggest concern is that market participants will place an excessive amount of confidence in results driven by new analytical techniques and not remember what we have learned from hard experience in the mortgage market: reliance on any analytical technique should be tempered by judgment.
4. With so few delinquencies in agency pools, some MBS investors have been hesitant to spend money on new data subscriptions or new tools. Do you think that will change? What sources of data or analytical methods hold the most promise?
We don’t wish for an economic downturn, but we do like to encourage our clients to be prepared prior to changes in economic conditions. Despite the relative calm in delinquencies and losses, we have seen firms expand their analytical capabilities as they enter new markets or prepare to meet changes in regulation (such as stress tests) or accounting (such as CECL). The use of these techniques has expanded the market’s use of probability-weighted scenarios, which provide insight not only into expected losses but also the risk of loss in stress scenarios.
One promising area for new data is the ability to combine performance data on mortgages with the credit bureau and other borrower data. For example, Fannie Mae and Equifax are able to link the CAS data and credit bureau data and can provide that to investors and others in a form that protects borrower confidential information. This represents a step forward from prior efforts to combine this data which relied on a statistical match.
5. How might impending demographic changes, inventory shortages and other headwinds impact the secondary market for mortgages?
Over the past few months, we have witnessed some slowing in home price appreciation. This trend may continue due to rising interest rates and reduced tax deductibility of mortgage interest expense. Home prices have also reached high levels due to the high cost of acquiring land and construction in the coastal markets. In addition, generally rising interest rates could lead to losses for mortgage investors and limit their appetite for additional assets.
Mortgage investments have several advantages relative to other asset classes. Improved underwriting since the financial crisis probably limits the credit risk of the mortgage market. In fact, we have seen much better performance of the enterprise-issued CRT bonds than similarly rated corporate bonds. Higher rates also reduce prepayment risk. It is possible that in the next few years, mortgages might provide a degree of stability for investors instead of being the source of market disruption.
Andrew Davidson is president of Andrew Davidson & Co., Inc., a New York firm specializing in the application of analytical tools to investment management, which he founded in 1992. He is a financial innovator and leader in the development of financial research and analytics. He has worked extensively on mortgage-backed securities product development, valuation and hedging.
Andrew Davidson & Co., Inc. turns mortgage data into investment insight. The firm created VECTORS® Analytics, a set of proprietary tools including the LoanDynamics Model for credit-sensitive mortgage securities, prepayment and option-adjusted spread (OAS) models for fixed-rate mortgages, adjustable-rate mortgages, collateralized mortgage obligations (CMOs), and asset-backed securities (ABS). Over 150 financial institutions depend on VECTORS® Analytics to help manage risk and value securities.
The company also provides consulting advice to financial institutions in the development and implementation of investment management and risk management strategies. They also work on a variety of fixed-income trading and valuation analyses. Customers of the firm include businesses of all sizes including many of the largest and most sophisticated financial institutions.
Andrew was instrumental in the creation of the Freddie Mac and Fannie Mae risk-sharing transactions: STACR and CAS. These transactions allow Freddie Mac and Fannie Mae to attract private capital to bear credit risk, even as they remain in government conservatorship. Andrew is also active in other dimensions of GSE reform and has testified before the Senate Banking Committee on multiple occasions. Andrew also helped establish the Structured Finance Industry Group and served on the Executive Committee at its inception.
For six years Andrew worked at Merrill Lynch, where he was a Managing Director in charge of a staff of 60 financial and system analysts. In this role, he produced research reports and sophisticated analytical tools including prepayment and option-adjusted spread models, portfolio analysis tools, and was also responsible for the development of trading and risk management systems for the mortgage desk covering ARMs, CMOs, pass-throughs, IOs/POs and OTC options.
Andrew was previously a financial analyst in Exxon’s Treasurer’s Department. He received an MBA in Finance at the University of Chicago and a BA in Mathematics and Physics at Harvard.
He is co-author of the books Mortgage Valuation Models: Embedded Options, Risk and Uncertainty; Securitization: Structuring and Investment Analysis; and Mortgage-Backed Securities: Investment Analysis and Valuation Techniques. He has contributed to The Handbook of Mortgage-Backed Securities and other publications.