What? Flood models are biased?
In my previous post, I explored the two earthquake metrics that help insurers understand the likelihood of an earthquake affecting a specific location. This post is about the two metrics that express the severity of property damage in the event an earthquake does occur.
Unlike the science behind earthquake forecasting, which has limited capabilities, the science behind the civil engineering that enables buildings and structures to withstand earthquakes is well developed. Insurance analytics and earthquake models leverage this engineering know-how to estimate potential damage (or vulnerability, to use the modeling term) for a given building.
The modeling of earthquake risk for property insurance is an inexact science. This is because the inner workings of the triggers are so obscure from direct observation; events of interest are relatively infrequent; and the time period between events is so long. These difficulties lead directly to the kind of innovation and creativity necessary to build models that live up to George Box’s tenet that “all models are wrong, but some are useful.”