Valuation of residential mortgage default and prepayment under stochastic house prices

This dissertation applies a valuation model for residential mortgages subject to exogenouslyspecified conditional probabilities of prepayment and default to data on subprime andAlt-A adjustable-rate mortgages. In conformity with the doubly stochastic framework,risk-neutral baseline hazard rates of prepayment and default in the model are dependenton stochastically evolving latent factors, driven by Brownian motions. Hazard processesalso incorporate observable variables reflecting evolution of interest rates and houseprices. The model is estimated on two subsets of the data: one includes early vintagesof non-prime ARMs, the other is comprised of 2005 and later vintages. Results of thecalibration of the two sets of model parameters to market prices suggest that whateverchanges in the unobserved borrowers behavior took place in the observation period wererelatively soon recognized by the market. Simulations show that for subprime borrowersthe effect of house price fluctuations on the probability of default is very significant.