This paper summarises the data collected through the 2002 Operational Risk Loss Data Collection Exercise (LDCE) launched by the Risk Management Group (RMG) of the Basel Committee on Banking Supervision in June 2002. The 2002 LDCE asked participating banks to provide information on individual operational losses during 2001, internal capital allocation for operational risk, expected operational losses, and a number of exposure indicators tied to specific business lines. The 2002 LDCE was an extension and refinement of two previous data collection exercises sponsored by the RMG, which focused on banks' internal capital allocations for operational risk and their overall operational risk loss experience during the period from 1998 to 2000.
The goals of this paper are to describe the results of the 2002 LDCE and to compare the data with the information collected by the RMG in its previous data collection efforts. As in the summary of the previous exercise, this paper focuses on the individual loss event data submitted by participating banks. To this end, the paper provides an analysis of the range of individual gross loss amounts and of the distribution of these losses across a set of standardised business lines and event types. It also provides an analysis of the information banks reported on insurance and other recoveries associated with these individual loss events. In both instances, the paper compares the information collected for 2001 with similar information for 2000.
Finally, the paper provides a brief examination of data collected on the share of economic capital that the participating banks allocated to operational risk, as well as their use of information on expected operational losses for pricing, reserving and expensing. The paper ends with a brief analysis of the comprehensiveness of 'exposure indicator' data provided by banks.
Background
The Committee's paper Working Paper on the Regulatory Treatment of Operational Risk published in September 2001 provides an overview of the proposed framework for the regulatory capital charge for operational risk. In the working paper, the Committee outlines proposals for the development of a capital charge to cover operational risk. In the proposals there are three approaches of increasing sophistication to assess the operational risk charge: the Basic Indicator Approach, the Standardised Approach and the Advanced Measurement Approaches (AMA).
As part of the second quantitative impact survey, the Committee conducted its first survey of operational risk data in May 2001. The data collected in that survey and in this current exercise will allow for the further calibration of the Basic Indicator and Standardised Approaches, and will inform the development of the AMA framework, in particular, resolving issues concerning the qualifying criteria for the AMA. The Committee envisages that these surveys will be part of an on-going data programme undertaken over the next few years to further refine the calibration of the operational risk charge.
The 2002 survey
The primary purpose of this survey is to collect granular (event-by-event) operational risk loss data to help the Committee determine the appropriate form and structure of the AMA. To facilitate the collection of comparable loss data at both the granular and aggregate levels across banks, the Committee is again using its detailed framework for classifying losses. In the framework, losses are classified in terms of a matrix comprising eight standard business lines and seven loss event categories. These seven event categories are then further divided into 20 sub-categories and the Committee would like to receive data on individual loss events classified at this second level of detail if available.
The Committee is also collecting information on six "exposure indicators" such as number of employees or total assets. The exposure indicator data serve two purposes. First, they are critical to the Committee's effort to aggregate loss data across banking institutions to arrive at an industry loss distribution. Second, the exposure indicators are necessary for banks and supervisors to relate historical loss experience to the current level of business activity. This information also enables banks and supervisors to determine separate frequency and severity distributions for the operational risk loss experience. Although indicators other than gross income are included in this survey, the Committee does not envision revisiting the use of gross income as the base for the Basic Indicator and Standardised Approaches.
The survey aims to collect data for the financial year 2001 and banks are asked to complete as much of the survey as possible on a best endeavours basis.
Changes to the previous survey
Although this survey is a repeat of an exercise carried out last year, there are a number of important changes to the scope and content of this year's survey. These include:
- Banks are no longer asked to provide operational risk loss data by `effect types',
- Banks are no longer asked to provide quarterly aggregated loss data,
- Banks are asked to provide data on expected as well as received recoveries,
- Banks are asked to indicate the internal threshold used for collecting loss data,
- Banks are asked to identify those losses arising from a `corporate centre' business,
- Banks are no longer asked to provide data on the value of transactions/deals/trades, the number of transactions/deals/trades, the standard deviation of transactions/deals/ trades, the number of accounts and the average balance of accounts,
- Banks are asked to provide component information on gross income,
- Simplified, structured Excel spreadsheets have been developed for completing this year's survey. These spreadsheets include a number of tests to check the consistency of data submitted and thus reduce the need for further enquiries after data are submitted. It is therefore important that banks do not change the design or structure of the questionnaire and check the messages of these tests.
The Committee appreciates that this exercise is a burden on banks. However, it believes, that its proposals will have increased accuracy and risk sensitivity where they are based on a sound quantitative foundation. The data requested are of fundamental importance to the development of the operational risk charge and the calibration of the overall capital framework, and to the extent that more accurate and complete data are received then the need for buffers or adjustments for uncertainty is reduced.