This article discusses how prepayment and default constitute competing risks in mortgage lending, provides examples of the importance of using a combined approach when evaluating the risk of whole loans and MBS, and concludes with practical implications of using the competing risks framework.

Though it may seem apt, the phrase "competing risks" in the title of this article does not refer to the annual budget battle between various risk management functions within large financial institutions. Rather, it is a framework for modeling the impact of separate causes for attrition. In the mortgage world, these are the separate, but interdependent, risks of prepayment and default. For prime mortgages (whole loans) and for mortgage-backed securities (MBS), prepayment risk has long dominated the issue of credit risk. Historically, in the secondary market, the three government-sponsored enterprises (GSEs) guaranteed the credit risk of most conforming mortgage loans, which represented the bulk of the primary market.

However, a recent issue of Inside Mortgage Finance (IMF, June 10, 2005) noted the following fact: "During the first three months of the year, non-prime lenders churned out an estimated $184 billion in new loans. Putting that into perspective, more than one out of every four loans--or 28.5%--of all new mortgages made during the first quarter of the year went to borrowers in the subprime and Alt A categories." Moreover, SMR Research Corporation estimates over $700 billion in junior liens outstanding at the end of 2004 (Home Equity Loans: 2005 Outlook). Therefore, loans with greater credit risk represent a significant and increasing portion of the primary mortgage market.
Advertisement

A large number of these loans, and the associated credit risk, are held on the balance sheets of large financial institutions. The IMF estimated that the 50 largest financial services holding companies held a combined $1.01 trillion in whole loans during that same first quarter of 2005. Moreover, according to Inside MBS & ABS (July 17, 2005), a weekly newsletter published by the IMF, Fannie Mae and Freddie Mac bought $212 billion in nonconforming MBS during 2004, assuming much of the credit risk for the underlying mortgages. (1) Thus, the increasing credit risk in the system is held by a large number of institutions, and even the GSEs now must learn to assess the greater credit risk from nonconforming products.

These recent developments increase the importance of default risk vis-a-vis prepayment risk for mortgage lenders, whether they are portfolio lenders or buyers of MBS. However, the presence of prepayment risk limits the applicability of traditional approaches to default modeling. The competing risks framework for modeling prepayment and default confers important advantages relative to these traditional approaches, particularly in the context of valuation, risk management, and capital allocation.

The Need for Hazard Models

As with other types of consumer assets, it is important to address the timing of the default event, and to account for static predictive variables. But mortgages are unique in offering the borrower an important and valuable option to prepay the loan early. Other consumer loans are prepayable, but only mortgages offer a significant financial reward for careful use of the option. This poses a particular challenge for default modelers. (2) Consumers frequently use the prepayment option when it is to their advantage. When interest rates reached 30-year lows in 2003, the monthly prepayment rate for prime mortgages reached nearly 7%. The average lives of mortgages vary enormously due to differing prepayment rates and lead to significantly different cumulative losses for mortgages with similar credit characteristics. As a result, building an accurate life-of-loan loss model for mortgages is very difficult.

Prepayment modelers take a different approach (as illustrated by the fact that the life-of-loan prepayment rate is a concept unheard of in the industry), using what are called hazard models. A hazard model is simply a model designed to predict the probability of attrition given that the subject has not yet left. For prepayment, this means predicting the probability of prepayment in a given month for all borrowers who have not yet prepaid. This methodology sees heavy use in prepayment modeling. (3) Hazard models for prepayment commonly include age and current rate levels as explanatory variables.

While the hazard modeling is well understood by investors and by Wall Street, technique has less commonly been applied to mortgage default. Hazard models, however, are frequently used in assessing the risk of default or bankruptcy for corporate bonds. (4) In the case of mortgages, a hazard model would predict the probability that the mortgage defaults in a particular month, given that it has not yet defaulted or prepaid. Such a model typically would include age, current house-price levels, and borrower FICO scores as explanatory variables.