Measuring and Managing Operational Risk in the Financial Sector: An Integrated Framework
University College London - Department of Computer Science
University of Liege - HEC Management School
HEC Management School - University of Liège; Maastricht University - Department of Finance; Gambit Financial Solutions
Deloitte Luxembourg; University of Liege - Economics, Business Administration and Social Sciences
February 27, 2005
This paper proposes a methodology to analyze the implications of the Advanced Measurement Approach (AMA) put forward by the Basel II Accord for the assessment of operational risk. We develop an integrated procedure for the construction of the distribution of aggregate losses, using internal and external data. It is illustrated on a 2x2 matrix of two selected business lines and two event types, drawn from a database of 3000 losses obtained from a large European banking institution. For each cell, the method calibrates three truncated distributions functions for the body of internal data, the tail of internal data, and external data. When the dependence structure between aggregate losses and the non-linear adjustment of external data are explicitly taken into account, the regulatory capital computed with the AMA method is substantially lower than with less sophisticated approaches, although the effect is not uniform. We then estimate the effects of operational risk management actions on bank profitability, through a measure of RAROC adapted to operational risk. The results suggest that substantial savings can be achieved through active management techniques, although the effect of a reduction of the frequency or severity of operational losses depends on the calibration of the aggregate loss distributions.
Number of Pages in PDF File: 33
Keywords: Basel 2, operational risk, extreme value theory, external data, RAROC
JEL Classification: G20, G21, G28working papers series
Date posted: May 7, 2005
© 2014 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo8 in 0.422 seconds