Reverse stress testing: Scenario design for macroprudential stress tests

We propose a systematic algorithmic reverse-stress testing methodology to create “worst case” scenarios for regulatory stress tests by accounting losses that arise from distressed portfolio liquidations. First, we derive the optimal bank response any given shock. Then, introduce an algorithm which systematically generates exploit key vulnerabilities in banks' holdings and thus maximize contagion despite apply our data of 2016 European Banking Authority (EBA) test, design worst case banks at time. Using spectral clustering techniques, group 10,000 worst-case into twelve geographically concentrated families. Our results show even though there is wide range different within these 12 families, each cluster tends affect same banks. An “Anna Karenina” principle emerges: Not all stressful are alike, but every scenario stresses These findings suggest precise specification not primal importance as long most vulnerable targeted sufficiently stressed. Finally, can be used uncover weakest links financial system thereby focus supervisory attention on these, building bridge between macroprudential microprudential tests.