A new study by the Bermuda Monetary Authority (BMA) on catastrophe modelling practices on the island has found a significant uptick in the proportion of insurers purchasing external catastrophe reinsurance.
The BMA asked insurers how their cat risk modelling reflects their reinsurance and retrocessional purchases, and found that the number of insurers purchasing little or no external cat reinsurance dropped from 20.0% of respondents in 2017 to 2.3% in 2021.
It concluded that the vast majority of insurers model cat risk by considering external reinsurance explicitly, either for some treaties or each separately.
In 2021, 93% of respondents explicitly calculated either some external reinsurance or all reinsurance treaties in their cat modelling.
Overall, the BMA concluded that modelling practices in Bermuda remain steady, with 88.73% of the cat risk exposure underwritten determined to be modellable using vendor cat models, and 98.47% of these risks having actually been modelled.
“With a relatively high concentration of catastrophe risk underwritten in Bermuda, a broad understanding of catastrophe modelling practices is central to the Bermuda Monetary Authority’s (Authority or BMA) supervisory framework,” said BMA Managing Director Ricardo Garcia.
“As insurers rely significantly upon internally built and external vendor models to assess catastrophe exposures, this information is also important to Bermuda insurers, other stakeholders and markets around the globe,” he continued.
In 2021, Bermuda legal entities’ average loading factor in the accumulation process increased to 9.3% compared to 6.6% in 2020 due to social inflation and the impact of climate change that is not yet reflected in the models.
For groups, the average loading factor stood at 8.8% in 2021, compared to 4.6% in 2020.
AIR and RMS were found to be the most frequently used modelling software tools, with RMS becoming the leading model for groups.
In-house modelling was utilised by 36.4% of legal entities and 31.6% of groups in 2021, and 34.4% of legal entities and 22.2% of groups reported using more than one model in their accumulations.
