Stress testing the German mortgage market

This paper presents a framework for estimating losses in the residential real estate mortgage portfolios of German banks. We develop an EL model where LGD estimates are based on current collateral values and PD dynamics are estimated using a structural PVAR approach. We confirm empirically that foreclosure rates are rising with the unemployment rate and are inversely related to house price inflation. Being consistent with our expectation that strategic defaults do not play a central role given the full personal liability of German households, the results give broad support for the double-trigger hypothesis of mortgage defaults. In order to analyse the possible credit losses stemming from residential mortgage lending we then use the model to run a top-down stress test and simulate losses on the individual bank level for the years from 2018 to 2020 for the whole German banking sector. Our results show that loss rates in the residential mortgage portfolios of German banks do increase significantly in an adverse economic environment. The estimated expected losses are widely distributed in the banking system leading, on average, to a 0.4 percentage points reduction in the CET1 ratio over the simulation period.