190 likes | 314 Vues
Alternative Models for Competing Risks of Mortgage Termination John Clapp School of Business Administration University of Connecticut and Yongheng Deng Lusk Center for Real Estate University of Southern California Prepared for AsRES-AREUEA Joint International Conference, Seoul, July 2002.
E N D
Alternative Models for Competing Risks of Mortgage Termination John Clapp School of Business Administration University of Connecticut and Yongheng Deng Lusk Center for Real Estate University of Southern California Prepared for AsRES-AREUEA Joint International Conference, Seoul, July 2002
Mortgage market trendsMortgage Debt Outstanding (MDO) and GDP
Mortgage Securities Transactions Have Grown Source: Federal Reserve Board
Competing Risks of Mortgage Termination • Mortgage Termination by Refinance • Green and Shoven (1986) • Hendershott and Van Order (1987) • Schwartz and Torous (1989) • Kau and Keenan (1995), Kau et al (1992)(1995) • Stanton (1994) • Archer, Ling and McGill (1996) • Caplin, Freeman and Tracy (1997) • Peristiani et al. (1997) • Deng, Quigley and Van Order (2000) • Downing, Stanton and Wallace (2001)
Competing Risks of Mortgage Termination • Mortgage Termination by Household Mobility • Quigley and Weinberg (1977) • Quigley (1987) • Brueckner (1992)(1994) • Clapp, Harding and LaCour-Little (2001) • Clapp, Goldberg, Harding and LaCour-Little (2001) • Pavlov (2001)
Competing Risks of Mortgage Termination • Mortgage Termination by Default • Foster and Van Order (1984) • Dunningham and Hendershott (1984) • Epperson, Kau, Keenan and Muller (9185) • Van Order and Zorn (1993) • Kau, Keenan and Kim (1993)(1994) • Lekkas, Quigley and Van Order (1994) • Vandell (1993)(1995) • Deng, Quigley and Van Order (1996) • Kau and Keenan (1999)
Alternative Models for Competing Risks of Mortgage Termination • Multinomial Logit Model (MNL) • Campbell and Dietrich (1983) • Cunningham and Capone (1990) • Arch, Ling and McGill (1996) • Clapp, Goldberg, Harding and LaCour-Little (2001) • Calhoun and Deng (2002)
Alternative Models for Competing Risks of Mortgage Termination • Cox Partial Likelihood Model (CPL) • Green and Shoven (1986) • Quigley, Van Order and Deng (1993) • Clapp, Goldberg, Harding and LaCour-Little (2001) • Pavlov (2001)
Alternative Models for Competing Risks of Mortgage Termination • Full Information Competing Risks Hazard Model (FICH) • Schwartz and Torous (1989) • Single risk model without unobserved heterogeneity • Deng, Quigley and Van Order (2000) • Mass-point mixture hazard model with competing risks • Deng and Quigley (2001) • Simulated Maximum Likelihood Estimation (SMLE) for competing risks hazard model with continuous distribution of unobserved heterogeneity
Our Plan • Extend the DQV (2000) two competing risks framework to three competing risks of mortgage termination by refinance, move and default • Examine the role of unobserved heterogeneity using a Mass-Point Mixture Multinomial Logit specification • Comparable to DQV (2000) Mass-Point Mixture Hazard Model with Competing Risks • Compare the performance of alternative models in terms of out-of-sample predictive ability
Discrete Cox Proportional Hazard Model • Cox Proportional Hazard Function • Cox Partial Likelihood Function
Log Likelihood Function ofMass-Point Mixture Multinomial Logit Model
Covariates • Market price of loan • Original loan balance • 15-year loan indicator • Points (estimated) • Original refinance indicator • Probability of negative equity > 90 percentile indicator • House price appreciation for age group < 40 • House price appreciation for age group > 40
Covariates • Unemployment rate • Borrower age • Minority indicator • Borrower income • Obligation ratio • High credit score indicator • Low credit score indicator
The Data • Transactions data from the California Market Data Cooperative, Inc (CMDC) • January 1993 through December 1998 • Detailed property information • Detailed loan information including • The origination date • Loan amount • Appraisal value • Street address of underlying property • Detailed borrower characteristics including • Credit score