With this subject being a hotter potato right now than a Jersey Royal locked in a sauna, in a tank top, in Bangkok, I've had a trawl through and found the following highlights:
- Most reasonable business uses of EC metrics appear to be applied or planned by respondents (pricing/underwriting decisions being the straggler)
- 40% of respondents said management understanding of EC is still limited - schematic on p9 showing the differences between 'sophisticated' Europe and 'savage' RoW hints at some kind of Solvency II dividend, though the results are not flattering across the board.
- Interesting schematic on implementation difficulties (p12), broken down by continent - data quality seems to have topped the list of implementation problems, which is no surprise I guess, but they neatly connect it with potential for over-reliance on expert judgement, simplifications and approximations to fill the gaps (all of which are to the detriment of a pure EC approach, at least in theory).
- Curve fitting is the majority-used approach (58%) to deliver model outputs quicker (i.e. 'lite' modelling), with replicating portfolios and LSMC less favoured
- Understanding around fungibility and dependency higlighted as areas for improvement
- Two-thirds still not allowing for sovereign debt risk in their EC calcs - I admire the persistency!
- Very interesting bit towards the back on effectively projecting EC, the holy grail for anyone in the ORSA space right now. While they loosely refer to the business planning horizon as "typically 3 years" (I've seen longer than that before breakfast, lads!), the point made is perfectly valid, namely that methodologies for this kind of projection are in their infancy.
- Finally, a nugget on Operational Risk Modelling (which I only touched on yesterday!), by some distance the least effective part of respondent's EC frameworks. They do note that 60% of EU respondents have moved to stochastic-based Op risk models with all bar one using expert judgement to calibrate them! They also bemoan the lack of credible data and subjectivity around cause/effect/latency of op risk events.
Also surprised that more reference to ratings agency demands wasn't made, particularly with that element seemingly influencing EC calibration points in the EU right now (hands up if you're at 1-in-2,000!)