The likely imposition by regulators of minimum standards for capital to cover 'other risks' has
been a driving force behind the recent interest in operational risk management. Much discussion
has been centered on the form of capital charges for other risks. At the same time major banks are
developing models to improve internal management of operational processes, new insurance
products for operational risks are being designed and there is growing interest in alternative risk
transfer, through OR-linked products.
The purpose of this paper is to introduce Bayesian belief networks (BBNs) and influence
diagrams for measuring and managing certain operational risks, such as transaction processing
risks and human risks. BBNs lend themselves to the causal modelling of operational processes: if
the causal factors can be identified, the Bayesian network will model the influences between
these factors and their contribution to the performance of the process. The ability to refine the
architecture and parameters of a BBN through back testing is explained, and the paper also
demonstrates the use of scenario analysis in a BBN to identify states that lead to maximum
operational losses.
Many thanks to Telia Weisman of Randloph Ivy Women's College for enthusiastic research
assistance.
ECE engineer school
