Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.574447
Title: A framework for modeling mortgage risk under abnormal market conditions
Author: Molina Utrilla, José Antonio
Awarding Body: University of Essex
Current Institution: University of Essex
Date of Award: 2012
Availability of Full Text:
Access through EThOS:
Abstract:
The abnormally high mortgage default rates that became apparent in late 2006 were not foreseen by standard forecasting methods in June 2005, when mortgage production in the US reached its peak. The effect of loose under- writing practices was masked by a consistently high house price appreciation between 2003 and 2005. This study provides strong evidence that the high US nonprime mortgage default rates were predictable in mid-2005 using his- torical data only available at the time. Based on a logistic regression and Markov chain framework, this research develops a suite of loan-level predictive models for mortgage delinquency, prepayment, and default rates. These models were implemented for new Fixed Rate Mortgage loans originated at the peak of the US mortgage is- suance, without making any assumptions on either actual data for later pe- riods or projected economic variables. The results show close agreement between actual default rates and predictions for a 5-year projected time win- dow starting in June 2005, when the high default rates of late 2006 were not foreseen. The early high prepayment rates from subprime borrowers over the 15 months prior to the crisis, the first sign of a dramatic increase iri defaults in case house prices dropped, as actually occurred, are also captured. If these predictions had been available in mid-2005, investors in Mortgage-Backed Se- curities would have demanded higher returns and lenders would have in turn tightened their underwriting practices at the right time.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.574447  DOI: Not available
Share: