Can the industry achieve the levels of automation reached by the financial industry in the modelling of risks? asks Tom Larsen
In discussions with the CAT modelling groups of property insurance companies, a question that often arises is "How will CAT model usage in the future differ from now?" This is a question that is key not only to model vendors to the property insurance industry, but also to reinsurers who are placed under continual pressure to maintain the bottom line even in the face of a soft market.
In order to effectively understand the future development of CAT modelling we must first gain a firm understanding of where we currently stand. Property CAT models were "adopted" in the years following Hurricane Andrew. CAT modelling is now a part of the underwriting and portfolio management processes of most insurers that have large exposure to losses. But it remains a very labour-intensive process of data transfer and interpretation, analysis and interpretation of CAT model results. This time-consuming process impedes decision-making and limits the ability to implement pricing and underwriting strategies.
Where do we look for examples of what can be done? The securities industry has used models for a long time to evaluate the risk levels of individual securities in their portfolios, and their portfolios as a whole. Each type of security may be subjected to more than one risk model to evaluate the risk associated with it. Regulation and competition have forced them to "mark to market" their portfolios on a daily basis. Achieving this level of modeling has required a significant amount of IT automation.
What, therefore, must the property insurance industry do to attain a level of automation and integration analogous to the banking and securities industries? Many of the necessary components are already forming. Industry data exchange standards are certainly a necessary component of achieving streamlined CAT modelling and the champions of standards are now appearing on the scene. Reducing the inefficiencies with uncertain data types is a key step towards achieving any level of integration. The final step is the formalisation of the processes in underwriting and accumulation to ensure that effective controls are in place.
Will the property insurance industry ever achieve the same level of automation that the financial industry has reached? The key limit seems to be the lack of liquidity in insurance contracts versus bonds or stocks. There will always be a need for a relationship-based understanding of particular cedants' requirements. Automation of CAT modelling will not completely change the practice of reinsurance underwriting but rather will complement it.