Pricing delinquent mortgages

Abstract

One of the most challenging tasks facing financial institutions is how much to mark down mortgage loans with delinquency events in the book of business. Since delinquent loans are not actively traded in the market, a modelling approach is needed for re-pricing these loans. Here, we show a two-step methodology for re-pricing delinquent loans. The first step uses the history of mortgage delinquencies to mark down the Fair Isaac Corporation Credit (FICO) score. The second step applies the mark down FICO score to a risk price model.

Notes

The article solely represents the authors' own perspectives and opinions.

¹Because foreclosures are lengthy dynamic processes, financial institutions use expected (rather than actual) losses in their financial report. Specifically, Generally Accepted Accounting Principles (GAAP) require companies to set aside money today for credit losses they expect to realize over the next 18–24 months.

²Risk-based pricing is an alignment of loan risk pricing with expected loan risk − charging a higher interest rate for higher risk (Freeman and Hamilton, 2002; Johnson, 2002; White, 2004).

³In 2006, the combined LTV on Alt-A purchase originations was 88%, with 55% taking out simultaneous seconds (piggybacks) at the time of purchase. Low/no documentation loans (stated income loans) represented 81% of total Alt-A purchase originations. IO and option Adjustable Rate Mortgage (ARM) loans represented approximately 62% of Alt-A purchase originations in 2006.

⁴The FICO score is based on a generic model of delinquencies that does not fully capture the impact of a mortgage delinquency on credit risk.

⁵The use of an early default model allows us to capture relative odds of default, while avoiding the complexities of a foreclosure model that takes years to occur.

⁶This parsimonious specification contrasts to the use of spatial effects that would involve modelling over 3400 municipalities (counties).

⁷The use of an early default model allows us to capture relative odds of default, while avoiding the complexities of a foreclosure model that takes years to occur.

⁸Variable interactions between FICO and past delinquencies are not statistically significant.