PROCEEDINGS May 15, 16, 17, 18, 2005 - Casualty Actuarial Society

ARCHITECTURE FOR RESIDENTIAL PROPERTY INSURANCE RATEMAKING 499is why loss cost (per policy) trends and earned rate trends areused to adjust the losses and premiums in each experience ratio.Accordingly, the de-trend factor for modeled cat losses isnecessary to state the expected cat losses on the same volumeof exposures as is underlying the approximate midpoint of eachexperience period. By nature, cat models produce losses givenin-force exposures as of a predetermined (but presumably recent)date. Due to run-time, data storage, and labor costs, it is impracticalto repeatedly simulate losses using in-force exposures fromseveral historical years. As an alternative, the selected annualchange in earned house-years from Exhibit 2 is raised to a negativepower representing the trend period between the in-forcedate used in the cat model and the midpoint of each experienceperiod to derive a de-trend factor. The factor is applied to thesingle modeled expected loss estimate to get the cat losses thatare loaded into each period shown on Exhibit 1. 10Note that the match between the attributes of excluded actualcat losses and modeled expected cat losses should be asclose as possible for actuarially efficient ratemaking. Claims departmentsare often responsible for coding individual claims as“catastrophic,” and there is generally no mandate for consistencywith the basis used for simulated cat losses. For example, if modeledhurricane losses reflect only landfalling hurricanes, but theclaims unit designates weak, bypassing tropical storms as thebasis for many “cat” claims received during a season, the excludedlosses are broader than the simulated losses that replacethem, making overall rate level indications inadequate. Actuariesshould be vigilant and proactive in setting parameters for cat losscoding with respect to:1. Event definitions (example: hurricane versus tropicalstorm)10 The de-trended cat losses are not then trended forward to the midpoint of the effectiveperiod because the modeled loss per exposure unit for cat perils is not inherentlyinflationary. Annual updates to models reflect the latest meteorological and scientificknowledge but not cost trends per dollar of value insured.