ABSTRACT
Insurers: Insurers typically keep summaries of insured properties and claims data. The data are proprietary and difficult to acquire, both for propriety as well as practical reasons. In the last decade, the data have moved from hard copy to electronic formats, which has modestly increased its availability. There are several potential sources of statistical bias inherent in insurance data, both for value exposed and loss amount. Insurers tend to select for target markets by income category, type of business, etc., and therefore their data may not reflect the general population of structures. Insurers do not show value exposed per se, but rather limit of liability, which can poorly reflect replacement cost. Extracting loss data is also challenging. Some insurance-claims databases exclude buildings with repair costs greater than zero, but less than the deductible, which can be substantial. Claim amount can differ significantly from actual earthquake-induced damage because of loss-adjustment practices and insurers’ public-image and other business considerations. Finally, insurers do not record shaking intensity; and while they do typically record an address, which can be used to find intensity from other sources, the address might be the billing address or the address of the main insured facility, rather than the site where the loss occurred.
