The PD-LGD Relation for US Mortgages

This thesis shows that while macro variables and default rates share common cycles for conventional US mortgages, a unique cycle is observed for loss given default, implying that the relation between the default rate and loss given default is weak for conventional US mortgages. The average loss given default across the US increases from 2002 until the end of the data set in 2014. Similar increases are observed across the U.S. Census Bureau’s defined regions, although the West region shows losses that seem impacted by the financial crisis of 2007-2008. The research finds evidence of a unique cycle for loss given default across the US as well as for specific regions, moreover this cycle is also present for uninsured mortgage loans with a high enough loan to value at origination ratio. The research uses a mixed-measurement dynamic factor model that applies a mixture of gaussians to capture the dynamics of a bimodal loss given default distribution. The finding of a unique loss given default cycle for conventional US mortgage loans is innovative, as previous research primarily focuses on corporate loans and finds a shared cycle between defaults and losses.