Modeling Fundamentals: AIR Industry Exposure Databases
Jan 26, 2021
Editor's Note: This article updates the one we published in 2016 on this topic, which was authored by Cheryl Hayes; it provides the most up-to-date information on how AIR's industry exposure databases (IEDs) are developed and maintained for all modeled countries, the numerous unique advantages they provide, and why the reliability of catastrophe model results depends so heavily upon them.
Industry exposure databases (IEDs) provide a foundation for all catastrophe model loss estimates. They contain counts of all insurable properties and their respective replacement values for a given country, along with information about the occupancy and the physical characteristics of the structures, such as construction type and height classification. Even information pertaining to standard industry policy conditions, such as limits and deductibles, is incorporated into a country's IED.
Benefits and Uses of IEDs
There are numerous benefits and uses of IEDs in catastrophe modeling. Estimates of potential losses from catastrophic events (such as hurricanes, earthquakes, floods, wildfires, or terrorist attacks) are generated based on IEDs. This information can be used by (re)insurance, broker, financial, corporate, and government clients to quantify and better understand catastrophic risk and the potential impacts on their business and assets.
Probabilistic loss distributions are derived from the IEDs, covering all regions and perils modeled by AIR. IED values are fundamental in calculating industry loss estimates for all event types, i.e., simulated events from a stochastic catalog, the “recreation” of historical events, or for actual events unfolding in real time.
Using IEDs with Touchstone Re and Touchstone
Through Touchstone ReTM, AIR's application for comprehensive catastrophe risk analysis using aggregate exposure data, companies can use IEDs to evaluate their loss potential. The total insured value of a company's books is compared to total industry-wide exposure in an IED to calculate a company's market share, which can then be used to prorate an industry loss to determine the company's share of that loss.
Touchstone Re industry loss estimates based on IED values can also be used to gain insight into the accuracy of modeled losses from Touchstone® , AIR's detailed, high-resolution risk management platform. Catastrophe loss estimates generated by detailed catastrophe modeling are only as good as the quality of the exposure data entered. If your market share is 1%, for example, you should expect something on the order of 1% of industry losses. A detailed analysis that indicates significantly lower or higher losses may suggest the need for further scrutiny of your company's catastrophe exposure data. Poor geocoding of risks, inaccurate replacement costs, and lack of information about occupancy and building characteristics could be contributing to inaccurate modeled estimates.
While models in Touchstone provide robust estimates of losses using detailed exposure data, there are instances where you may have exposure data with unknown building characteristics. Losses can vary greatly depending on location, which is significant when modeling buildings with a combination of unknown primary characteristics, such as construction and height. For this type of exposure, IEDs can be used to estimate damage functions; the model applies an “unknown” damage function based on country-specific construction characteristics from the IED in some cases, and in others the AIR model significantly enhances modeling of these unknown combinations by considering local building attributes at the regional, county or province, or CRESTA level.
In addition to making use of unknown damage functions, Touchstone users can also obtain better loss estimates using a feature that allows you to disaggregate coarse-resolution portfolio exposure data to a highly detailed level in line with the spatial distribution of AIR’s IEDs.
Benchmarking exposure to catastrophe risk can also be accomplished with IEDs. For example, you may be undervaluing your exposures—exposing your company to greater losses than your exposure assumptions would reveal—which can be determined by comparing your portfolio with IED industry data (available in IED documentation) or by comparing your portfolio to the aggregate-level industry exposures (available for export from Touchstone Re).
Clients who want to take a more in-depth look at exposure in the United States can license AIR’s U.S. IED to perform these types of analyses at a much higher resolution. Additional characteristics such as risk counts by detailed occupancy, binned height classes, and construction types are included at the ZIP Code level when this is licensed by clients. Licensees can use this IED to: analyze potential growth strategies; disaggregate exposure data to a highly detailed level; perform sensitivity analyses; conduct real-time analytics; validate models; and utilize exposure data for market share analysis.
How Does AIR Develop Our IEDs?
AIR uses a comprehensive, bottom-up approach to develop IEDs. The process begins by obtaining detailed data on risk counts, building characteristics, and construction costs from many sources. Depending on the country of interest, the information may be provided in various languages; additional steps may be necessary to translate the data before it can be used. The resolution and vintage of the data can also vary by source. As a result, the task of compiling and analyzing these diverse data sets can be extremely time-consuming. The preparation of an IED requires skill and experience working with demographic data, along with an understanding of structural engineering, cost estimation, and economics.
Given the importance of IEDs in catastrophe modeling, AIR devotes considerable resources to their creation. Each IED is developed by a dedicated multi-disciplinary team comprising engineers, economists, GIS specialists, construction cost estimators, insurance professionals, and other specialties. To date, AIR has created building inventory databases for more than 110 countries around the world.
Around the World in IEDs
The first IED developed by AIR was for the Unites States. We began compiling data in 1987, and full updates are generally completed every two years. Residential and commercial risk counts are obtained each year from a private data provider. The current estimates of residential risk counts are based on census data, which is the primary source of residential risk information. Actual counts of commercial establishments are provided each year for commercial risks, and cost indices are used to update replacement costs. The U.S. IED is developed at a very high geographic resolution—up to 90-meter grid resolution—thanks to highly detailed input data, which are particularly critical in modeling perils that tend to have smaller footprints and cause more localized damage.
Each country presents its own challenges, and it can be particularly difficult to create IEDs for developing countries. Limited access to data or poor data quality has been problematic in such regions as South America and parts of Southeast Asia.
Lack of current data can also be an issue. Census data, for example, provides a good source of information for IEDs, but full surveys are often conducted infrequently (about every 10 years). It is up to the members of the Exposures team to apply their skill and expertise to transform these data sets and bring in proxy data, as necessary, to model the risk counts and attributes to build robust IEDs for these countries.
Validation: A Critical Step in the IED Development Process
Validation during AIR’s IED development process is the step when all components used to build an IED are thoroughly reviewed and scrutinized to ensure accuracy. In contrast to the bottom-up approach used to build an IED, a top-down approach is used for validation.
Once the replacement values in an IED are derived, they are validated against client data values and country-specific economic data on fixed assets and other macroeconomic variables. Using industry reference materials and discussions with clients, we also validate the policy terms and conditions we use in our IEDs to make sure they align with the standard market practices in a given country by peril. In addition, our Exposures team spends a significant amount of time validating individual large industrial facility risks using sources such as industry reports and location-specific news and financials. Finally, when catastrophic events take place, we perform detailed damage surveys to verify that the construction splits on the ground are properly reflected in our IEDs.
Keeping the IEDs Current
Once an IED is developed, it must be continually updated to accurately reflect current values. For example, minor growth in replacement values in the U.S. can result in an increase of a few trillion dollars in total value. Updates are also particularly important in developing countries with rapidly expanding economies, where the annual change can be quite significant.
The replacement values contained in an IED are developed using a rebuild cost approach in which risk counts (number of properties), building characteristics, and costs act as the main drivers. Due to the dynamic nature of these elements, new information on each must be gathered for each update. An increase in total industry insured value of residential properties, for example, can be caused by multiple factors such as an increase in risk counts, adjustments in housing sizes, and updated costs.
With each update AIR collects the best available data for each country and uses advanced technology and techniques to build high-quality IEDs. AIR’s latest modeling enhancements are captured in the recently released updated exposure for the Caribbean—a high-resolution, 90-meter grid IED was created for each of the 29 modeled countries.
As we think about future IED updates for countries that have limited data sets available or have a large exposure base, we are constantly seeking out the latest technologies to improve the quality of our IEDs. Machine learning is one of the areas we are taking close look at. In China for example, we are exploring the use of machine learning in conjunction with satellite imagery to locate clusters of high-rise buildings.
Due to the time-consuming and labor-intensive nature of IED creation, full updates for countries outside of the U.S. coincide with major model updates. To ensure that modeled industry losses are kept current, AIR has developed Industry Exposure Indexes that reflect changes in asset values at a country level across all business lines. Indexes for individual countries are updated periodically and can be found in the AIR Industry Exposure Indexes for Select Countries document (login required). Industry exposure indexes are also used when AIR issues ALERT™ industry loss estimates for actual events in countries for which an index has been published.
IEDs for More Accurate Modeled Risk Estimates
The value of industry exposure databases to the modeling process cannot be overemphasized, from providing the foundation for industry loss estimates to a multitude of additional applications. Thus, investment in data sets and the experts skilled at constructing IEDs for all modeled countries is a critically important attribute of any catastrophe modeling company and of primary importance at AIR.
Companies can take advantage of this detailed exposure data in their risk assessment for more accurate industry loss estimates. IEDs can also be used to validate building characteristics to more accurately assess a portfolio’s risk. By leveraging AIR’s IEDs, you can make more accurate modeled risk estimates.