Ed. note: Reprinted with permission from GfK GeoMarketing.
GfK GeoMarketing: You created a windstorm model for many European countries. How does this model work?
RMS: The RMS Europe Windstorm Model helps protect households and other insured policy holders from the risk of their insurance companies becoming bankrupt if a major windstorm sweeps across Europe. While some severe windstorms have been experienced in recent times, such as windstorms Daria, Vivian and Wiebke in 1990, Kyrill in 2007, and Emma in 2008, science and history show that there is a chance of even more severe storms than those occurring in the future. The model was designed to extrapolate beyond history to the worst possible storm, and capture all possible events that can happen, and how frequently these might occur.
The model used scientific principles, including meteorological forecasting models and global climate models, combined with a dense network of wind observation data, to estimate the probability of damage from a full range of possible windstorms at any location across 15 European countries. To achieve this, RMS simulated four main model components:
- a set of around 30,000 possible windstorm events that could strike Europe
- a realistic meteorological model that estimates the peak winds experienced at each location throughout the duration of each of these events, which accounts for terrain and local environmental impacts
- functions that relate the severity of the wind to the damage that would be sustained by building types of all different ages, shapes and sizes
- the financial losses that would be sustained by policyholders and insurers
GfK GeoMarketing: How is it set up? What is the level of detail?
RMS: The key value of the model is that it provides a highly informed estimate of the damage potential across an individual insurer's entire portfolio of risks. For very big windstorms impacting a large swath of Europe, an insurer may not have sufficient funds to pay out all its policyholders unless it knows its full exposure. The model is flexible enough to work with the information available. Where the risk is highest and changes the most rapidly, for example around cities, or at coastlines where differences in topography and the environment can cause sudden changes in wind speeds, the model calculates each event's peak winds for every 1 km² grid cell. To optimize computing times, RMS uses larger cells of up to 10 km² in areas where the risk does not change much spatially. If users know the location and physical characteristics of the individual buildings in a particular grid cell, the model can calculate expected losses for individual buildings. Conversely, the model can also use aggregated data. For example, the insurer may only know the total number of properties it insures in a particular post code or county, along with the type of usage, for example, residential or commercial, and the total value at risk.
GfK GeoMarketing: Which data is fed into the model?
RMS: Model users - typically insurance companies (as well as reinsurers and brokers) - must input the location, value and characteristics of the buildings insured, along with the insurance policy terms, such as any excess the policyholder is liable for, before the insurance policy kicks in. To understand market-wide losses, users can input values from the RMS Industry Exposure Database.
GfK GeoMarketing data was a key resource RMS used to create its Industry Exposure Database. The Industry Exposure Database is a detailed snapshot of all the insured property in every postal code in Europe. RMS used this snapshot during the development of the windstorm model - for example, to validate the model by comparing total market losses predicted using the RMS Industry Exposure Database with the losses reported at the time of the event. The GfK data, which covers all the European countries in the RMS model, combines several key metrics of interest with consistent vintage, resolution and format geospatial reference data. RMS used demographic data on population and households, combined with consistent postal code boundary maps, to understand the location of residential properties. This is particularly useful in countries where the statistical office does not provide data with accompanying geospatial information, or to break out coarser-resolution data (for example, the number of single-family dwellings by region) to a higher resolution. RMS used purchasing power to refine its valuation models. For example, in Paris purchasing power was used to differentiate between wealthy districts with high rebuild costs per square meter and poorer districts with poorer quality, cheaper housing. In addition, the German data on number of businesses was helpful in understanding the geospatial distribution of insured non-residential property in Germany.
GfK GeoMarketing: How do RMS and its customers use the windstorm model?
RMS: Model users only need to know the location and characteristics of property at risk, as well as the sums insured. Once this information is entered, the model can conduct analyses. Users with additional information - for example, about the type of construction, policy limits and deductibles - can also input this data to assess the impact of these characteristics. Users can choose from a range of analysis types, including scenario events, such as "worst-case" analysis or "what-if" analysis – for example, "What if windstorm XX happened tomorrow?” The real value of the model, however, stems from probabilistic analyses. Though these take more time, they give a full distribution of potential losses and the probability across an insurance company's entire book of business. This allows an insurer to determine the probability of sustaining losses of particular amounts in any given year - for example, by calculating that while there is a 5% chance of losses totaling €300m, there is a 1% chance of a loss totaling €700m - which allows it to manage the amount of funds it keeps available for paying potential losses.