Catastrophe models failed to accurately capture the destruction following the 2010‘s Maulé earthquake to the point that carriers have become unsure of the adequacy of their reinsurance coverage, according to a report from Aon Benfield.
“A year after the event, the issue of model miss is evident,” the report titled Chile: One Year On.
According to the report released Friday, the original modeled loss estimates following the February 27, 2010, 8.8 magnitude quake ranged from 5 percent to 13 percent of key zone aggregates (KZA); a metric used by carriers to determine the amount of reinsurance coverage.
“This meant that for some of the event footprints, the loss would have exceeded the regulatory minimum reinsurance requirement of 10.5% of KZA,” the report states.
However, actual losses were much smaller and contributed to significant “model miss” following the disaster, the report said. Modeled losses were nearly 100 percent higher than actual losses for personal lines. The estimated losses for commercial lines were between 10 percent to 100 percent higher than claims.
Aon Benfield points out that the experience in Chile was “reversed” from U.S. hurricane model miss, where models have traditionally underestimated actual losses.
“Going forward the focus will be on the model calibration, particularly in light of the actual claims data,” the report said. “Model changes may be needed if the Chilean regulator decides to change the current fixed percentage rules to a modeled based approach.”
Although the report does not detail the reason for the models’ failure to accurately estimate losses in Chile, it says that it was likely driven by “modeling firms underestimating the quality of construction and how it would perform during an earthquake.”
Building codes in Chile are often compared to those in the United States, albeit they are focused on preventing a total collapse rather than non-structural damage. Recent reports state that the Chilean government is considering further strengthening codes for new buildings and revising the system of ground classification to account for the flexibility of the soil.
Aon Benfield says that catastrophe modeling firms will need to narrow the gap between estimated and actual losses in order to “build insurers’ confidence in using models as a risk assessment tool.
“To that end, the users of these models will need to feel comfortable with the vulnerability and hazard components of the models, including time dependency, post loss amplification and industry exposure databases,” the report stated.
