Comparing Solvency II Standard Scenarios for Windstorms with Catastrophe Model Outcomes – Updated Study: Part I
With the generalized use of catastrophe models to measure the natural catastrophe exposure of insurance portfolios, the outcomes of these models have more and more influence in the determination of reinsurance needs. With the introduction of the Solvency II regime, the decision on reinsurance purchase should also be an integral part of a company’s risk management process. Specifically, an important consideration is the impact of reinsurance contracts on the Solvency Capital Ratio, a key decision metric of the risk management process. This process is not always easy when the probable maximum losses (PMLs) derived by the cat models differ from the standard European scenarios under Solvency II for calculation of the Solvency Capital Requirement for cat risk (SCRCat).
To better understand the possible differences that can emerge between the cat modeling and standard scenario results, we have updated our analysis that measures the industry-wide impact of the wind catastrophe scenarios provided in the standard formula of Solvency II. Since our first study, which was published in December 2010 (Quantitative Impact Study (QIS)5 Scenarios are Within Range of Industry Models), two important changes have taken place:
- Some of the commercial catastrophe models have been updated and there have been major new releases for modeling of European Wind. The outcomes of the analyses of the new versions of the models are sometimes very different from the outcomes from the previous versions. These have led to differences in the definition of the 200 year SCRCat results provided by these tools under a (partial) internal model framework.
- The European Insurance and Occupational Pensions Authority (EIOPA) is working on the Level II Implementing Measures and some countries have been developing new national QIS exercises. Thus far, the proposed new templates for the standard scenarios have not impacted the definition of the different wind scenarios which were proposed by QIS5.
Guy Carpenter’s analysis uses the last version of the European insurance industry exposure database portion of the PERILS AG Industry Exposure and Loss Database (PERILS Industry Database). This new exposure database was published in April 2012 as part of the annual updates that PERILS makes available to its subscribers.
The Standard Formula Scenarios
The process of defining the QIS5 natural catastrophe scenarios was supported by an industry-wide task force consisting of major industry players (including Guy Carpenter) as well as regulatory authorities. Following the QIS5 exercise, the EIOPA cat working group’s timeline was extended. It has been focusing on two main initiatives - a review of existing scenarios in light of the QIS5 outcomes and proposing new scenarios for the perils and countries that were not (sufficiently) covered by the QIS5 scenarios. No major adjustment has been made to the QIS5 scenarios for European Wind.
In our analysis, we first ran the standard scenarios for each country and for Europe on the PERILS Industry Database. These results are summarized in the following graph.
The diversified wind standard scenario for the countries covered by the European insurance industry exposure database portion of the PERILS Industry Database amounts to EUR39 billion. The exposure is dominated by the United Kingdom, France and Germany, where the standard scenario for wind shows amounts in excess of EUR10 billion. Two Nordic countries, Norway and Sweden, have been recently added to the PERILS database and their national standard scenario amounts to EUR1.4 billion and EUR2 billion respectively. The diversification benefit amounts to EUR19.9 billion or 34 percent of the sum of the eleven national scenarios.
Standard Scenarios versus Industry Loss Models
The second step in our analysis was to compare the loss figures computed with the PERILS Industry Database under the standard scenarios to the 1-in-200 year loss figures in the last available version of the three largest cat model vendors. In this analysis, all figures were computed based on gross figures, including impact of deductibles and limits, on an occurrence basis.
Table 1 and Figure 2 illustrate the comparison of the loss projections computed with the cat vendors’ models and with the standard formula for European wind. Our new study shows conclusions similar to those of the December 2010 study. While most outcomes in these new model versions are falling within the range of the three selected vendor models, some important divergences with the standard scenarios are observed.
In most cases, the standard scenarios lie in the interval of results from the vendor models or are very comparable to it. Generally speaking, each model tends to be higher or lower than the standard scenarios for most countries. The three cat models show lower results in Germany and The Netherlands and higher results in Ireland. Some isolated large differences between the standard scenario and one of the three models are observed for some countries. On a European combined basis, while one model shows a result very close to the outcome from the standard formula, the other two models are lower. The diversification effect, while substantial in all four approaches, is distinctly lower in the standard scenarios. The differences in diversification benefits have an important impact in the comparison of the European combined results.
Using the PERILS Industry Database, 200-year losses under QIS5 standard scenarios by country are shown next to the corresponding losses from the major three vendor models.