Topic: Contributed on Measuring and forecasting default risk Title: Modelling loss severity for residential mortgage loans: A three-step selection approach Authors:  Hung Do - University of Technology Sydney (Australia) [presenting]
Harald Scheule - University of Technology Sydney (Australia)
Daniel Rosch - University of Regensburg (Germany)
Abstract: We develop a new framework to model the loss given default (LGD) of residential mortgage loans, one of the most important asset classes in banks. According to Finance perspective, defaulted loans with zero-LGD are intuitively more relevant to the liquidity constrains of the borrowers (such as, divorce, demotion or loss of job); whereas, the non-zero LGD loans often relates to negative equity issues. Therefore, we suggest treating zero- and non-zero LGD loans distinctively. We propose a three-step selection approach with a joint probability framework among default, cure (i.e, zero-LGD) and loss severity information. Our model better fits the Finance perspective as well as the bimodal nature of the LGD distribution, which shows a massive concentration on zero. An application of our framework on the U.S residential mortgage loans during the Global Financial Crisis shows interesting new evidence that: (i) foreclosed residential mortgage loans associated with higher credit quality borrowers are less likely to be fully recovered; (ii) regional foreclosure contagion effect exists in an essence that a rise in local foreclosure rate increases the probability of default, lowers the probability of cure and worsens the non-zero LGD; (iii) house price under distress of the foreclosed property exacerbates the non-zero LGDs locally.