Coruscation

Modeling Residential Mortgage Termination and Severity
Using Loan Level Data

Three essays on modeling residential mortgages.

Chapter 1 presents and estimates a new model of loss given
default using a new dataset of prime and subprime mortgages. The
model combines option theory proxies with information on the loan
contract and the cash flow position of the borrower. The results
suggest that severity on subprime and adjustable rate mortgages are
similar to losses on fixed rate prime loans, but that investor owned
properties have significantly higher losses than owner occupied
houses. The results also suggest systemic overappraisals on refinanced
loans.

Chapter 2 uses option pricing methodology to value the prepayment and
default options associated with a residential mortgage, if house
prices are mean reverting.

Numerical solutions compare the results from the mean reverting house
price model to the results from a model where house prices follow a
geometric Brownian motion process.

The main contributions are:

(1) the value of the implicit rent (service flow) is derived as a
function of the house price process instead of assumed to be constant,
as in prior research;

(2) the mean reverting model has additional factors that may help
forecast mortgage termination; and

(3) the house price process is shown to have a significant effect on
the value of a mortgage over a wide range of parameter values.

Chapter 3 presents a modeling framework for residential mortgages that
has separate models for each loan payment status (Current, 30 Days
Late, 60 Days Late, 90+ Days Late, in Foreclosure, in REO, or Paid
Off). It is shown that several classes of traditional mortgage
prepayment and default models are restricted forms of this model, and
that the restrictions are rejected empirically.