Exposure Data

Why Exposure Data Is Critical

The hazard module tells the model where events occur and how intense they are. But a severe earthquake in an uninhabited desert causes no insured loss. To translate hazard into loss, the model needs an accurate description of the inventory — the portfolio of properties at risk. The quality of exposure data is, arguably, the single most important factor in the accuracy of a cat model output. As practitioners often say: garbage in, garbage out.

What Exposure Data Contains

For aggregate industry-level analysis, modellers maintain annually updated databases from governmental and private sources, containing estimates of total property exposures within the modelled region at the postal code level. For individual company portfolio analysis, exposure data is submitted by the client and typically contains:

• Location — the most important parameter. Properties must be geocoded (assigned latitude/longitude) from street addresses, ZIP codes, or other descriptors
• Construction type — wood frame, masonry, reinforced concrete, steel frame
• Occupancy — residential, commercial, industrial, agricultural
• Number of storeys — height is a key determinant of structural response
• Age of construction — reflects the building code in force at time of construction
• Insured replacement value — the maximum financial exposure
• Policy terms — deductibles, limits, sub-limits, coinsurance

Why Construction Type Matters

Building damage is primarily a function of construction type, and different construction types perform very differently under different perils. Masonry buildings, for example, typically perform poorly when subjected to violent earthquake ground shaking (unreinforced masonry is particularly vulnerable) but perform quite well against hurricane winds. Wood-frame buildings fare relatively well in earthquakes but can suffer severe wind and flood damage. Engineered buildings — those designed by structural engineers to current codes — typically outperform non-engineered buildings regardless of peril.

Regional Differences in Building Practice

Exposure databases must account for regional differences in construction practice and building code adoption and enforcement. A wood-frame house built in California in 2010 has been designed to a modern seismic code with specific lateral bracing requirements. The same construction type built in 1965 in the same location would likely be far more vulnerable. Similarly, a concrete building in Turkey built before modern seismic codes may be extremely vulnerable to earthquake despite being a "concrete" structure.

Contents and Time Element

Catastrophe models estimate damage not just to buildings but to their contents and indirect losses. Contents damage is typically a function of both occupancy class (which indicates what kinds of contents are inside) and structural damage to the building envelope. Time element losses — also called business interruption or additional living expenses — capture the cost of not being able to use the building during repairs.

Data Quality and Working with Clients

When estimating losses on individual insurance company portfolios, modellers work closely with clients to identify missing or erroneous data and to test for reasonability. The more detailed and accurate the information provided, the more detailed and reliable the model output. For particularly important or valuable buildings, a site-specific analysis may be appropriate — involving on-site engineering inspections and actual design documents rather than class-level assumptions.