The challenge of assessing the capital, necessary to protect an organization against exposure to operational risk losses is discussed in this paper (operational risk itself is defined as the risk of loss arising from inadequate or failed internal processes, people and systems or from external events). The evolutionary nature of operational risk modelling to establish capital charges is recognized emphasizing the importance of capturing tail behavior. Challenges surrounding the quantification of operational risk particularly those associated to sparse data are addressed with modern statistical methodology including nonparametric smoothing techniques with a particular view to comparison with Extreme Value Theory (EVT). The credibility approach employed supports analysis from pooled data across business lines on a dataset from an internationally active insurance company. The approach has the potential to be applied more generally, for example where data might be pooled across risk types or where a combination of internal company losses and publicly reported (external) data is used.
There is an increasing interest in financial services companies in identifying loss
distributions associated with operational risks, driven by both regulatory considerations
and also in recognition of the greater importance placed on operational risk management.
Building on 1988, Basel Capital Accord, the Basel Committee on Banking Supervision
published; International Convergence of Capital Measurement and Capital Standards1, in
June, 2004. This document addressed the challenge of how much capital is necessary to
protect an organization against unexpected losses, and established the need for an explicit
charge for the exposure to operational risk losses. Operational risk itself is defined as the
risk of loss arising from inadequate or failed internal processes, people and systems or from
external events. The framework also recognized the evolutionary nature of operational risk
modelling emphasizing the importance of capturing tail events.
This paper discusses the challenges surrounding the quantification of operational risk
and lack of suitable data and it also focuses on tail behaviour. These challenges are
addressed with modern statistical methodologies including nonparametric smoothing
techniques. One of the commonly applied techniques in operational risk is Extreme Value
Theory (EVT), which offers a broadly accepted methodology for estimating the tail of a
distribution, see Embrecht, Kliippelberg and Mikosch (1999) for a detailed mathematical
treatment, and also Embrecht (2000), Reiss and Thomas (2000) and Coles (2001). |
