“Contributions to Modeling of Operational Risk in Banks”

Advisor: Svetlozar T. Rachev

University of California at Santa Barbara, 2006

Abstract:

Part I of the dissertation reviews the Basel II Capital Accord and the regulatory capital requirements for operational risk.

Part II of the dissertation addresses statistical and probabilistic assessment of operational risk. Value-at-Risk and Conditional Value-at-Risk are taken as the measure of the buffer capital. Losses are modeled in an actuarial-type compound Cox process framework. A variety of issues is addressed; among the topics discussed are the following. (1) We provide empirical evidence that the intensity factor follows a very specific non-homogeneous form. (2) Internal operational loss databases suffer from reporting bias; practitioners often neglect this issue. We provide theoretical and empirical justification that this bias leads to severe underestimation of the capital charge, and propose methodologies to evaluate the exact information loss and incorporate it into the operational risk model. (3) “Low frequency/high severity” events are successfully captured by fitting variations of the alpha-Stable distribution to the loss severity data. (4) An innovative EDF-based goodness-of-fit test is designed to evaluate the performance of the loss distributions in the upper quantiles that largely determine the amount of the risk capital. (5) Finally, methodologies for robust modeling of operational risk are addressed.