Stress-testing and credit risk revisited: a shipping sector application

Conventional stress-testing in credit risk management may considerably underestimate economic losses associated with the most negative scenarios. In this paper, we show that in order to properly stress-test credit risk, we need to derive initially the default correlation among assets or companies. Then a risk measure needs to be applied to the stressed default rate distribution, in order to obtain default rates attached to the most negative scenarios. We find that the application of both the stressed default correlation and the risk measure deliver considerable deviations from the conventional approach. To illustrate our approach we use 192 publicly listed shipping corporations, over 2000-2016, and show that conventional stress-testing tools may underestimate losses under certain conditions. Furthermore, we show that the Vasicek (1987) model assumption of independence, between the systemic factor and the default correlation may not hold in certain cases. This assumption may actually be the reason for the likelihood of underestimation of credit risk capital requirements as applied in the context of the Basel internal ratings-based (IRB) approach.