ECO-SLV /05/2010

Transcription

1 Please insert your comments in the table below, and send it to in word format. Re ference Comment General comment We think that the draft CAT technical specifications are generally a good step in the right direction. In general we support the idea that the standardised scenario method is the default method and the factor approach only has to be applied on an exceptional basis. However, when neither of these methods is appropriate, undertakings should be able to use personalised scenarios subject to supervisory approval. QIS5 may be used to test these personalised scenarios. It is essential that there is complete and clear description of the approaches to be used under the Cat Risks modules in QIS5. As they currently stand, the QIS5 Draft Technical Specifications cannot be appropriately assessed as most of the catastrophe scenarios underlying the proposed calibration factors for each peril have not been specified. In addition, the wording is often unclear and some of the formulas cannot be properly understood. Parts of the formulas are missing and different acronyms are used inconsistently. Further work needs to be done in order to ensure that the specifications are clear and the approaches proposed are consistent between each other and with other parts of the specifications. To this extent, it is essential to specify 1/8

2 how the cat risk modules for Non-Life and Health articulate with the Group SCR section and how in particular diversification is taken into account at group level for cat risks. Also, it is unclear why the calculation methods for man-made Cat risks are different for each Peril/LoB. If there is no rationale for these differences then harmonisation should be sought. The calculation of the cat capital charge on a multi-lob level is not in line with how some undertakings manage their cat risks. The business mix in the LoB Fire can be extremely heterogeneous. This heterogeneity increases further if one combines different LoBs such as Fire, MAT and Motor. There are further difficulties as the exposure measure is totally different. For example in Germany, the generally used exposure measure for Motor is vehicle years which cannot be combined with insured sum as exposure measure for Fire and MAT. Regarding Natural Catastrophes, we support a regional/geographical segmentation, but we do not support a segmentation based only on models and calculations done by a commercial Cat model provider. Furthermore, it is not possible to assess how another zone-system would influence the parameters proposed for the Cresta zones since the underlying scenarios used to derive these parameters have not yet been made available. Regarding Man-made Cat risk, CEIOPS final advice proposed a change to the factor method in that it should be applied gross of reinsurance and the resultant losses put through undertakings reinsurance programmes. The factors were calibrated gross of reinsurance and, as such, saw a significant increase on the factors proposed in QIS4. The QIS5 draft specification applies these gross factors to premiums net of reinsurance. This is a clear mismatch or mistake. We understand there is much more work needed to prepare an approach of a Cat model fitting all Cat risks in all countries in Europe in a standard approach. CEA stands ready to cooperate with the Cat TF especially in the light of no direct insurer being involved. Section 1 2/8

3 Para It needs to be made clear for which countries storm surge risk has been included in the windstorm peril factors. Strom surge is not a peril to which Germany is exposed and as such it would be inappropriate to allow for this peril within the windstorm factors for Germany First bullet point: it is unclear how the scope of the scenarios has been defined to include the French Dom Tom. Fourth bullet point: the technical simplifications should consider a calculation based per line of business, not o nly a reallocation mechanism It should be pointed out that the application of the factor based method is a very exceptional case It should be pointed out that the application of the factor based method is a very exceptional case The specifications say that undertakings writing material amounts of non-proportional reinsurance or with material exposures outside the EU are expected to seek internal model approval. The specifications make no provision for such undertakings, which are required to use one of the two methods, both of which are explicitly inappropriate for them. We suggest a third method, of allowing undertaking specific calculation of non-life catastrophe risk. This could be allowed under two conditions: 1. Methods 1 and 2 can be demonstrated to be inappropriate. 2. The undertaking specific approach includes detailed validation of assumptions. In practice, this approach would require supervisory approval as with other USP, although this would not be appropriate for inclusion in QIS5. A similar comment to that for USP should be inserted, to the effect that if the data / validation standards are not currently met but are expected to be in force by Solvency II implementation then the numbers can be included for QIS5 purposes. This approach would ensure that the standard formula is available to those for whom methods 1 and 2 are inappropriate We generally support this approach. Nevertheless, it is not clear how this approach fits with the other sections of the cat risks 3/8

4 9.68 specifications. In addition, we don t have clear understanding of what happens if reinsurance is placed at several levels of segmentation (per LoB, per peril, joint reinsurance for several perils etc.). Is it allowed to apply reinsurance at each level of segmentation? We suggest the technical specifications to include other standard reinsurance structures, such as stop loss The reinstatement premium should be only taken into account if the risk mitigation of the reinstatement is also taken into account. Otherwise, the reinstatement increases the capital charge whereas a (less prudent) reinsurance contract without reinstatement results in a lower capital charge Formulae should be revised. It is not clear which correlation factor is meant in which formula There might be better / other possible zonings besides cresta Regional Segmentation: We think that a regional segmentation of covers exposed to Nat CAT risks is appropriate. But CRESTA should not be the only possibility of such a regional segmentation of total insured values. Comment on CRESTA zones: In Germany the undertakings use for some covers, finer zones than the broader regional zones. It is not possible to assess how another zone-system would influence F, Q and AGG in the formulae, in particular since the underlying scenarios used to derive the parameters for the Cresta zones have not been made available More explanations about the two scenarios are needed. How can a multi-peril reinsurance structure be taken into account? Is it possible to apply t he reinsurance on each peril independently? What if the reinsurance is only valid for parts of the windstorm portfolio? The naming conventions need reconsideration: in SCR.9.77 the calculation of the gross Cat_windstorm is described; then in SCR.9.78 the calculation of Net_Cat_windstorm is described, but the output is again Cat_windstorm? 9.79 There might be better / other possible zonings besides cresta /8

5 9.81 There might be better / other possible zonings besides cresta How can a multi-peril reinsurance structure be taken into account? Is it possible to apply t he reinsurance on each peril independently? What if the reinsurance is only valid for parts of the earthquake portfolio? Please reconsider the nomenclature: in SCR the calculation of the gross Cat_flood is described, then in SCR.9.83 the calculation of Net_Cat_flood is described, but the output is again Cat_flood? 9.85 We recommend a harmonisation of calculation methods between perils/lobs concerning man- made. We stand ready to make available to you proposals developed by national associations We don t understand why the description scenario approach is used. Scenarios are described at the beginning, but the calculation is a factor approach It is not taken into account that these man- made scenarios, as described in SCR.9.91, are very local events and thus the diversification / concentration of the portfolio is a material parameter in the determination of the man- made cat risk. This is not considered in the factor approach (market factor * sum insured). The approach implies an overestimation of the risk for a well diversified portfolio and an underestimation for a highly concentrated portfolio. What is the definition of the LSR = largest single risk? Does it mean the policy with the highest insured sum in the portfolio or does it also take into account the concentration of several policies within a certain radius? For a large and well diversified portfolio, the LSR might be the upper bound for this man-made risk; it might be underestimating the true underlying risk in a small but concentrated portfolio if the LSR does not include the combination of several policies Again: please reconsider the nomenclature. Is Cat_fire gross or net of reinsurance? 9.95 We don t understand why the description scenario approach is used. Scenarios are described at the beginning, but the calculation is a factor approach. The output is already mentioned to be net of reinsurance in SCR.9.94, but again risk mitigation adjustments should be applied due to SCR.9.95? Compare to SCR.9.106: output is mentioned to be gross 5/8

6 9.98 What is the difference between VY_country and LIM_country? What does this mean? Is the W_country calculated market wide (country wide) or per single undertaking? We don t understand the rationale for this parameter The definition of the factor F_country is not clear. How is a proportion to unlimited cover defined? Formula is not clear. Is there a solution of this equation available? We do not understand the relation between VY and TVY. How should this formula work in case that VY_country is unlimited? We don t consider that it s appropriate to mix marine property and marine liability. Shouldn t it be specified whether a property or a liability claim is mentioned? SI_Lt and SI_Lo: what to do in case of unlimited cover? How needs the calculation described in SCR be taken into account? We acknowledge that the risk on large exposures is material as evidenced by the practice of credit insurers purchasing excess-of-loss reinsurance. However, since credit insurers manage their top exposures prudently, the three largest exposures will likely have a higher than average credit quality (invest ment grade or comparable). Consequently, default of the three largest exposures within a 12- month horizon appears inconsistent with the 99.5% confidence level. It is required to add the magnitude of the increase of risk associated with reinsurance recovery considered in letter (II), to the extent this risk has not already been considered in counterparty default risk SCR. The counterparty default risk is alre ady explicitly included in the LGD calculation for reinsurers, therefore this specification appears to be unnecessary. There does not seem to be a mechanism to avoid double-counting to ensure that the risk of default of large exposures is not already considered in the calibration of the premium/reserve risk sub- modules Calculation of the dampening mechanism of SCR CAT_recession_net is unclear and possibly very complex in comparison to the result. There does not seem to be a mechanism to avoid double-counting to ensure that the risk of failures in a recession is not already considered in the calibration of the premium/reserve risk sub- modules. 6/8

7 A new dampening mechanism is proposed without any formal consultation regarding the reason for such a mechanism. We suggest there is need for a consultation regarding the perception of (pro)cyclicality risks and how they should be addressed The 10 scenarios are not included, therefore we cannot comment. Regarding the assumption of EUR 10 million retention amount after reinsurance recovery, we would expect this amount to be company-specific and not the same for all companies, regardless of size or reinsurance arrangement. We would be interested in a rationale for this CEIOPS final advice proposed a change to the factor method in that it should be applied gross of re insurance and the resultant losses put through undertakings reinsurance programmes. The factors were calibrated gross of reinsurance and, as such, saw a significant increase on the factors proposed in QIS4. The QIS5 draft specification applies these gross factors to premiums net of reinsurance. This is a clear mismatch or mistake. The solution is either to apply the factors gross of reinsurance and let undertakings put the results through their reinsurance programmes (as per CEIOPS proposals) or to produc e factors net of reinsurance. Given the time constraints, the only net factors available are probably the QIS4 factors. We would propose using the QIS4 factors The calculation of the method 2-gross loss in the calibration paper was not understood. The resulting figures could not be reproduced using a parallel GDV model. There are concerns regarding the derivation of the method 2-factors (mean?, median?, weighted?). 8.4 There is a natural limit to health related costs in general. In particular with respect to pandemics the extent of costs is naturally limited by the capacity of the health systems to deal with increased demand for health services following a pandemic. Also in some cases, governments will step in for parts of the costs of pandemics as it was observed during the H1N1 pandemic. These should be taken into account by the Cat Risk module for Health. Please refer to our comments on NL Cat Risk where relevant. Eg. See our comments on reinstatement premium for reinsurance in the net capital cha rge calculation in the NL Cat Risk section. SCR Values of Ip (Insurance penetration for product type and by member state) seems totally incorrect for Italy, at least as 7/8

8 and J.2 Annex regards Income protection and Medical expenses insurance. 8/8