This article presents a reduced-form model that contains frailty factors to predict mortgage default and develops a novel framework to model systematic risk of mortgages. We match default rates along multiple dimensions by extending the generalized autoregressive score (GAS) models. Our data consist of commercial mortgages in the U.S. retail market, ranging from 1997 to 2013, and contain more than 2 million records. Our most complex model incorporates 15 frailty factors. Nevertheless, the estimation process only takes two minutes with a standard desktop computer. Many competing models require simulations and are, in comparison, time-consuming when a large dataset is used.
Note: Research papers posted on ResearchGate, including any findings, may differ from the final version chosen for publication in academic journals.