Academic journal article Journal of Financial Management & Analysis

Scenario Analysis for Modelling Operational Losses in the Absence of Data : The Spanish Bank in Perspective*

Academic journal article Journal of Financial Management & Analysis

Scenario Analysis for Modelling Operational Losses in the Absence of Data : The Spanish Bank in Perspective*

Article excerpt

Introduction

In the last few years, bank industry has suffered from important losses due to operational failures. With this respect, the Basel Committee on Banking Supervision (henceforth the Committee)1 published, in June 2004, the New Capital Accord (Basel II)2, one of whose main novelty was the inclusion of operational risk capital requirements. In the case of Spain, the transposition of the European Directive on Basel II3 has been recently done by the Circular 3/2008 on Minimum Capital Requirements of the Bank of Spain. This new regulatory framework has encouraged the Spanish banking sector to evolve up to sophisticated techniques for managing operational risk more effectively4. For this purpose, the Committee (2001) proposes three main approaches for calculating the capital requirements; ranked from lower to higher degree of sophistication and sensitivity to such risk, these are: the Basic Indicator Approach (BIA), the Standardized Approach (SA) and the Advanced Measurement Approach (AMA). In turn, the AMA models comprise three alternative methodologies: the Internal Measurement Approach (IMA), Scorecards and the Loss Distribution Approach (LDA). According to the Committee, the LDA seems to be the most suitable methodology for estimating the Capital at Risk (CaR). Inherited of the actuarial discipline, the main objective of this model is to obtain the aggregate loss distribution from which inferring the 99.9th percentile, named as the Operational Value at Risk (OpVaR). The LDA is based on historical operational losses for which we should model the corresponding distributions: the severity, or monetary amount of the loss; and, the frequency with which the event is repeated during a specified period of time; that is, the probability that the event may occurs. With this aim, the Committee (2006; 152)5 explicitly states that:

internally generated operational risk measures used for regulatory capital purposes must be based on a minimum five-year observation period of internal loss data, whether the internal loss data are used directly to build the loss measure or to validate it.

On the other hand, McNeil and Saladin' suggested that at least 25 exceedances beyond the threshold should be the minimum data set required for modelling with robustness; otherwise the estimates from little data may be unreliable.

In practice, this is the main shortcoming that Spanish Banks are facing since they don't have enough data collected in their Internal Operational Loss Database (IOLD) to apply the LDA rigorously. Further, Guillen et al.k also considered the under-reporting phenomenon; which consists in ignoring or not identifying particular losses generated by operational failures, such as, for example, small losses with high frequency that are not computed when the entity calculates the capital charge, although their aggregation could embody a serious threat to the solvency of the entity. For this reason, the Committee (2006; 153-154) recommends to complete the IOLD with the utilization of external databases. However, there may be cases where OpVaR99.9 per cent based primarily on both internal and external loss event data even would be unreliable for certain event types, characterized by heavy-tailed loss distribution and a few observations. At this point, the Scenario Analysis is a practical tool to offset such inconvenience.

In this paper, we use a Spanish Saving Bank's IOLD with three years collected operational events and 1 7,936 observations in total. We focussed on the estimation of the OpVaR99.9% by using the Scenario Analysis when lacking enough data, such as in the case of the Internal Fraud* event type, for which there is only one observation recorded. Moreover, we conduct a stress analysis on CaR with regard to the shape and scale parameters which characterize the Internal Fraud loss distribution. As a result, we point out the importance of developing the Scenario Analysis to complete the IOLD in order to obtain stronger CaR estimates for operational events with low frequency and high severity. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.