Natural catastrophes and terrorism can cripple the financial viability of an organization. Hurricane Andrew, in addition to causing more than $16 billion in insured damage, left at least 11 insurers insolvent in 1992. More recently, in 2012, post-tropical cyclone Sandy caused an estimated $20 billion in insured losses and put countless businesses in financial jeopardy as a result of storm surge damage and sustained power outages.
Fortunately, these sorts of occurrences are rare. But it is exactly their rarity that makes estimating losses from—and preparing for—future catastrophes so difficult. What’s more, the usefulness of the loss data that do exist is limited because of the constantly changing landscape of insured properties. Case in point: if Hurricane Andrew were to recur today, it would cost insurers nearly $60 billion. Companies looking to make better business decisions and manage their catastrophe risk more effectively simply cannot rely on scarce historical loss data to project future losses.
AIR Worldwide pioneered the catastrophe modeling industry to address these challenges, creating the tools and technologies that changed how people think about catastrophic risk. Today, catastrophe modeling is standard practice in the insurance industry and is being increasingly adopted by other segments.
How AIR Models Work?
AIR models are based on sophisticated simulation methods and powerful computer programs that capture how catastrophes—both natural and man-made—behave and impact the built environment. AIR scientists and engineers combine simulations of the natural occurrence patterns and characteristics of hurricanes, tornadoes, severe winter storms, earthquakes, and other catastrophes, with information on property values, construction types, and occupancy classes. Model output provides information concerning the potential for large losses before they occur so companies can prepare for their financial impact.
The catastrophe modeling framework illustrated here applies to all AIR models.
Large catalogs of simulated catastrophes are generated, representing the entire spectrum of plausible events. These catalogs answer the questions: Where are future events likely to occur? How large or severe are they likely to be? And how frequently are they likely to occur? For each simulated event, the model calculates the intensity at each location.
Measures of intensity—wind speed, ground shaking, flood depth—are then applied to highly detailed information about the properties that are exposed to them. Estimates of physical damage are translated into estimates of monetary damage.
For each simulated event, insured losses are calculated by applying policy conditions (deductibles and limits, for example) to the total damage estimates.
How Our Models Are Used
Catastrophe models provide detailed output from which various measures of loss potential and risk can be derived. One example is the average annual loss (AAL), which refers to the loss that can be expected to occur per year, on average, over a period of many years. Another important output is the exceedance probability (EP) curve, which reveals the probability that a loss of any given size (or greater) will occur in the coming year.
Today, catastrophe model output is the basis for understanding and quantifying catastrophe risk. It is the “currency” by which risk is priced, transferred and traded. AIR modeling is used extensively for pricing, risk selection and underwriting, loss mitigation activities, reinsurance decision-making and overall portfolio management. But it is not just the insurance industry that looks to AIR for help. Applications of the technology have broadened to serve the needs of corporate risk managers, government agencies, investors, hedge funds and other financial institutions, and a wide variety of other stakeholders exposed to catastrophe risk.
All Models Are Not Created Equal
While catastrophe models begin with the same historical data, different assumptions used in the model development can lead to differences in model output. To ensure that final model results are both realistic and robust, AIR builds its models from the ground up, validating each component independently. Critically, we also validate the models from the top down to ensure that final model results make sense.