Insurdata study reveals scale of exposure data challenge

Insurdata, the award-winning insurtech firm which specialises in the augmentation of peril-specific exposure and risk data via its Exposure Engine, has today released the findings of an exposure data resolution study for a sample portfolio of US-based insured properties.

The findings revealed that of the 4,500 locations assessed in the report, over 40 percent (over 1,800 locations) were displaced by more than 10 metres, with 10 percent of the location data inaccurate by between 500-1,000 metres. The combined results showed an average displacement of 183 metres for the entire data set, with the largest displacement being 33 kilometres.

Additional analysis conducted by Insurdata at the individual company level has revealed that the net impact on modelled loss estimates of correcting location latitudes and longitudes can have a material effect on Average Annualised Loss estimates.

Commenting on the degree of displacement evident in the data, Paul Burgess, Head of Client Development at Insurdata, said: “It is important to consider this level of displacement in the context of standard accumulation zones. If a (re)insurer is applying a 200 metre accumulation zone for its terrorism exposure, for example, our study reveals that at a minimum 6 percent of the risks will fall outside of the accumulation zones they are believed to be in, while locations with smaller displacements could also be outside of the zone perimeters.”

The study also analysed the data sets to assess first-floor elevation (FFE) information against standard model assumptions regularly applied to assets where no FFE data is available. The findings showed that the FFE measurements for the properties were much lower than default modelled assumptions for over 90 percent of locations in the study.

According to Jason Futers, CEO of Insurdata: “The results from this particular data set show that there is a strong likelihood that the modelled loss estimates for flood will be underestimated for these particular locations. Flood exposure is one of the most data intensive risks in the (re)insurance market. Without key attribute data such as FFE the value of applying high-resolution models to such perils will be significantly reduced.”

He continued: “The misrepresentation of location data and a lack of detailed building parameter information can have a significant detrimental impact on the ability of (re)insurance companies to accurately assess exposure levels from the individual risk through to the property portfolio. Advances in technology now mean we can generate property-specific data at a much more granular level for every asset in a portfolio in a quick, efficient and accurate manner. Companies must therefore take steps to capitalise on this to improve risk selection and pricing and reduce portfolio volatility.”

The analysis was based on a sample set of approximately 4,500 insured property assets. The location data, which was sourced from a variety of re/insurance entities, was for US-wide risks. The geocoded property data, which included building and zip-code level information, was augmented using Insurdata’s Exposure Engine with updated building centroid information and additional building attribute data, including FFE and building perimeter dimensions.

“The industry suffers from a sense of institutional amnesia as an industry - we operate a renewal business, we see the same risks year after year but tend to forget all of the enhancement we’ve made to all of the risk information each time we look at the risks”
Paul Nunn — Head of Catastrophe Modelling at SCOR