Methodology
PERILS collects its data directly from insurance companies underwriting business in the covered territories. Data provision is based on a contractual agreement (the Data Provider Agreement) between PERILS and each of the data-providing insurance companies.
The data provided include exposure data (sums insured) by CRESTA zone and by country, property premium data by country, and event loss data by CRESTA zone and by country. This company data is made anonymous upon receipt and is tested for quality and completeness using standardized data quality and completeness checks. The data is then added to the data which has already been accepted within the identical aggregation units in the PERILS database. The original raw company data is deleted at this stage in compliance with applicable antitrust and competition laws.
Aggregated company data within the identical aggregation units is extrapolated to industry-level (i.e. market-level) using published market property premium information per country. The latter is broken down into individual aggregation units using population data and other proxy data. Company property premium data per country is broken down into individual aggregation units using average rates as derived from provided sums insured and premium data. The relationship between market premium and company premium then gives the market coverage per aggregation unit. The latter is used to extrapolate event loss data per aggregation unit to industry-level (i.e. market-level). The calculated industry event loss is subsequently tested against information from national authorities and insurance industry sources, as well as meteorological and other scientific data. This ensures the highest possible degree of data quality and realism.
Flowchart illustrating the PERILS methodology. After data collection from insurance companies, data is made anonymous and aggregated within the identical aggregation units followed by extrapolation to industry level.
It is important to note that the reverse engineering of PERILS industry data is not possible, i.e. the data providing sources cannot be reconstructed and their anonymity is assured.