Abstract
Operational risk modelling helps the risk managers to better anticipate operational risk and hence it supports more efficient risk management. There are several techniques and methodological tools developed to fit frequency and severity models including the already-mentioned EVT, Bayesian inference, dynamic Bayesian networks, and expectation maximisation algorithms.
When modelling operational risk, other methods that change the number of researched data of operational risk events are used. The first one are the robust statistic methods used Chernobai and Ratchev (2007) that exclude outliers from a data sample.
On the other hand, a stress-testing method, what we analyzed in this paper, adds more data to a data sample and is widely used by financial institutions.