In Belgium, flood insurance coverage for residential risks, commonly known as 'simple risks' with a maximum sum insured of 1.11 million, has been under discussion for the last two decades. Especially in the aftermath of the catastrophic river floods of the 1990s, the future of the Fonds des Calamites (national disaster fund), created by the Belgian government to pay for natural catastrophe losses, was called into question over its claims handling and speed of claims payments.
There were a number of proposals for alternative insurance solutions and legislation, ranging from voluntary to fully compulsory cover, sometimes also including cover for other natural perils such as earthquake and landslide.
In 2003, the government passed a law mandating compulsory flood coverage, but limited it to high risk, flood prone zones around the large rivers.
This law requires that the three regions of the Belgian Federal State, Flanders, Wallonia and Brussels, define these high risk zones, but for technical reasons, this has not yet happened. Consequently, the law is not yet applicable. Assuralia, the Belgian insurance industry association, has consistently opposed this law because the risk would be inadequately distributed; only residents in high risk properties would be obliged to buy cover.
Assuralia is continuing to lobby for alternatives that would cover as large a group of the population as possible, but does not expect further developments this year. In the meantime, some 10 or 12 companies have begun offering their own flood insurance. These products vary from company to company. Some are included in the household policy. Other insurers offer flood cover as an addition to the standard policy subject to an additional premium.
This lack of consistency between companies in the way flood cover is offered and priced suggested a need for a recognised, market wide approach or tool to help insurance companies assess risks and price policies. Such a tool is particularly valuable when insurers want to transfer catastrophe level flood risks to the reinsurance market.
Guy Carpenter, therefore, sponsored the development of a Belgium river flood model by the catastrophe modeller, Risk Management Solutions (RMS).
Because such river flood models are built at very high resolution, they can be used for a variety of business-related purposes including pricing, underwriting and portfolio management, as well as defining common metrics that can be exchanged between these players for flood reinsurance pricing.
In developing this model, RMS was able to draw upon its expertise in the area of river flood modelling developed through years of experience creating its river flood model for the UK. The Belgian flood model will be available to both the insurance and reinsurance markets in autumn 2004.
Collaboration between Guy Carpenter's Brussels office and RMS focused mainly on finding organisations who could deliver suitable hydrological information, gathering meteorological and statistical data, and interpreting local insurance standards and historical information from the Fonds de Calamites.
Because the most recent available information was integrated into the model, it is a reliable tool to assess the Belgium flood exposure of insurance companies. During the development phase several test runs were made for interested insurance companies in order to calibrate and fine tune the model. In addition to a stochastic event set, historical as-if claims amounts from events during the 1990s were discussed.
Building the model
The model is comprised of four principal components. These components:
1. Define the set of stochastic or simulated rainfall events.
2. Calculate a measure of the flood hazard at each location, in this case, the depth of the flood water or the intensity of the rainfall.
3. Determine damage under the policies provided through an assessment of the vulnerability of the portfolio.
4. Calculate losses from a number of relevant financial perspectives.
1. The stochastic rainfall event set
The model is fully probabilistic: it uses sophisticated statistical techniques to create an efficient, stochastic set of rainfall events. These events are generated from 500,000 simulated years of history, based on and calibrated against the historical record of flooding events, using hydrological flow and rainfall data from the Flanders and Wallonia regions of Belgium. The final event set contains over 3,200 possible winter and summer events to represent the full spectrum of floods likely to affect insured properties in Belgium.
2. Flood hazard modeling
The purpose of the hazard calculation component of the model is to determine an appropriate measurement for the hazard. Overall, the aim is to develop a picture of regional and local risks, as well as the variability, particularly for high risk areas.
In areas of Belgium defined as within the major flood plains, the model calculates the likely depth of the flood water. The flow in each major river in flood is propagated across the floodplain using a hydrologic engineering model using a 20m digital terrain model (DTM) from the Belgian National Geographical Institute (IGN).
In areas outside the flood plains, flooding results from the rapid response of minor rivers, sheet-flow and drainage to intense rainfall, and the model uses rainfall intensity as a measure of the flood hazard in these situations. In both cases, the hazard is calculated for each location for each event in the stochastic event set. The model also takes into consideration the system of flood defences in place in Belgium and, incorporates the probability that each could fail or be overtopped for each event in the event set.
3. Vulnerability modelling
In the vulnerability component of the model, RMS developed an understanding of building response to river flooding. Vulnerability functions, also called damage curves, for different buildings/structures and contents, and the resulting loss of use, relate damage to a given flood depth or rainfall intensity. These functions relate the damage that occurs on the major floodplains to flood depth and to rainfall intensity outside the main flood plains. To determine these functions, RMS used its experience of vulnerability modelling for other catastrophe models, in particular for the UK flood model, technical literature on the subject, detailed analyses of historical claims data from the Fonds de Calamites and results from RMS flood surveys.
4. Calculation of flood related losses
This element of the model converts the vulnerability function into insured loss. It takes into account deductibles, limits and other insurance conditions, such as new for old contents replacement, to calculate the financial impact of each simulated event.
One of the principal outputs of the model is an exceedance probability (EP) curve that represents the probability that the maximum loss to the insurer for its portfolio from flood will exceed various monetary amounts in the course of a year.
Guy Carpenter sponsored a number of studies for the main Belgian insurance companies, which helped RMS test the model and obtain feedback from the individual companies about the results. The results of these studies were well received by the market and were in line with most companies' expectations.
Other natural perils
In addition to flood, Belgium is exposed to other natural perils, such as windstorm. So the risk management process of insurance companies in general and the structuring of reinsurance coverage in particular should be based on a holistic modelling approach for all these natural catastrophes.
Correlation between different perils and also between the different European countries must be considered, especially for the large European groups present in the Belgium market. Scientific advancements and rapid changing technology also make a continuing development of the cat modelling software a key issue.
Eddy Vanbeneden works with Guy Carpenter Instrat in Belgium, www.guycarp.com Fathia Grandjean is RMS relationship manager in Paris with responsibility for continental Europe. www.rms.com