CEIOPS issued additional guidance around the standard formula for calculating capital requirements in respect of operational risk late last year.
Why was a new OpRisk formula needed?
The original formula for OpRisk proposed in QIS4 was widely condemned. Complaints included being too simplistic, being insensitive to risk (and basely primarily on business size) and the impossibility of calibrating to 99.5% in a meaningful way. CEIOPS accepts most of this criticism, but counters by reminding stakeholders that the aim of the standard formula is partly about being simple.
A more serious problem is that in comparison against companies’ own internal models, the standard formula produced results lower than companies’ own assessment. Median internal model requirements for OpRisk were 133% of the standard formula and 13 out of 16 countries reported higher requirements under their insurers’ internal models.
One of the aims of the standard formula is to be slightly conservative to provide an incentive for insurers to develop their internal models. Clearly this objective is not being achieved.
Current OpRisk recommendation for Solvency II
CEIOPS has issued final (they’re calling it final anyway) level 2 guidance on OpRisk requirements under the standard formula.
Although there are a few detailed differences (around negative components, for example) but the most significant change for most insurers will relate to the change in parameters. In many cases the changes are close to doubling of the parameters. This will significantly increase capital requirements for many insurers.
|Parameter name||New Factors||Old Factors from QIS4|
|BSCR cap life||30%||30%|
|BSCR cap non-life||30%||30%|
QIS5 is planned, so presumably the financial implications of the new recommendations will be tested.
It’s still clear to most that OpRisk is particularly poorly suited to Pillar 1 and purely quantitative requirements. While the updated formula wont have won over many of the loudest critics, it does better match capital requirements to those companies were getting from their own models.
Best solution is still a company-tailored internal model combining actual loss data collected, supplemented by subjective frequency/severity assessments mapped to density functions using work-shopping techniques.
Other relevant CEIOPS documents on Operational Risk
Earlier suggested OpRisk paper (superseded by November version) (pdf)
Note from the Actuarial Society of South Africa outlining recommendations for use in conjunction with PGN104 (December 2009):