Response to CP05/3, Strengthening Capital Standards (QQ 134-145, Operational Risk)
from John Thirlwell (Former Director of the British Bankers' Association and the Operational Risk Research Forum)
135. No comment
137. Transitional introduction is almost certainly essential. Of more significance are the beta factors being used within the TSA. The Bank of Italy published an interesting paper last July (Moscadelli, 2004) which appeared to throw considerable doubt on (a) the calibration of the betas and (b) the proposed business line hierarchy. It would be helpful to understand the FSA's views on this and whether the betas are sacrosanct. At the very least, of course, no beta should be higher than the alpha agreed for the BIA.
138. No comment
139. If the betas remain as currently proposed, a firm may have a higher capital charge under the TSA compared with the BIA depending on the business lines they are in and the proportion of their income derived from them. (See A137 above.) This is intuitively wrong and hardly incentivises firms to "move along the spectrum of approaches".
The introduction of the negative income problem is yet another manifestation of the crudeness of the STA calculation and the fact that income has little correlation with risk.
On the balance of the necessary evils offered, I should prefer the Directive approach, since I am concerned that netting may also provide opportunity and encouragement for arbitrage.
140. Guidance on this aspect would be helpful.
141. Paragraph 11.8 highlights the critical challenges to firms if operational risk management is to be truly embedded in a firm's culture. Although, in general, I prefer guidance to rules, I believe that to provide further guidance at this stage beyond that contained in this section, would dilute a powerful message and point firms towards a tick-list rather than force them to engage in a real internal debate about the place of risk management in their organisation.
142. Reading paragraph 11.13, I have considerable sympathy with the industry experts with whom you have shared these thoughts. The sub-text in the paragraph is that data will be complete (third bullet) and wholly verifiable (first bullet). Whilst I can understand the FSA's wish for high standards, these have to be put in the context of both reality and, as the industry experts have commented, the requirements of other regulators. It is the nature of the 'non-loss event data' (i.e. self-assessment and scenario analysis) that much of it will be subjective. Even KRI's, whose correlation with future operational risk experience is not established at this stage, can legitimately include measures which reflect subjective (including binary) assessments. These views can be assessed, but that is far from saying that they can be "verified". As to completeness, it is an inescapable fact that loss event data will not be complete, possibly to a material degree. How far this invalidates the use to which firms put it is a matter of judgement for firms and their supervisors. I should therefore be concerned to see guidance which is based on a misconception of the realities of operation risk management until these matters have been further discussed and, indeed, used over a period of time.
144. Paragraph 11.23 already describes some of the methods by which firms "recognise" expected loss and also points to the fact that EL, within operational risk, means something different from its use in market risk and credit risk. In operational risk terms, operational risk being largely about business risks, it might fairly be described as the 'cost of doing business'. I therefore welcome an approach which allows firms to indicate what that means to them and how they assess and evaluate this, and explain and discuss this with their supervisors.
145. The key issue is to understand that to chase for a 99.9% confidence level is to chase after a shadow and possibly to bring operational risk into disrepute. The Basel Accord helpfully includes, at paragraph 669(f), recognition of elements and factors which render the 99.9% target as being unattainable. It is disappointing that the proposed Directive offers no such recognition. In a recent paper examining the AMA (Chapelle and others, 2005), a number of Benelux academics involved in business and management have suggested that the capital charge for the bank data they have used indicates that "the capital charge . . . should lie within a 20% two-sided interval from our point estimate with a 90% probability." This is a long way from 99.9% confidence. There will be considerable subjectivity involved in even a 95% estimate, let alone that involved in the 'boot-strapping' suggested in the FSA paper (11.25). In the circumstances, I am in complete agreement with your decision not to provide high-level guidance at this stage.
Chapelle and others,
2005 - Chapelle, A, Crama, Y, Hubner, G, Peters, J-P, January 2005. measuring and managing operational risk in the financial sector: An integrated framework. (Research financially supported by the National Bank of Belgium)
Moscadelli, 2004 - Moscadelli, M, July 2004. The modelling of operational risk: experience with the analysis of the data collected by the Basel Committee, Temi di discussione del Servizio Studi, Number 517, Banca d'Italia.