Within the insurance industry, understanding the location of a property is vital to accurately assess and price the risk.
For flood, this geospatial intelligence is especially important as the intensity of the hazard varies significantly at a small spatial scale. Some properties can be unaffected while their neighbours across the street can be left uninhabitable for months.
While advanced hydraulic models can capture the patterns of water flow and likely flood depths, modelling the depth of water within a property and the resultant damage to both the building structure and homeowner's content is highly uncertain. With high resolution flood models, understanding whether a property floods or not is key to accurately pricing the premium, calculating portfolio level maximum losses and identifying future claims.
How do insurers locate risk?
Insurers rely on geocoders to associate an address with a geographic coordinate that can be used by the catastrophe model. The accuracy of the result depends on the completeness of the address data and the sophistication of the geocoder's underlying datasets. The result could be matched to building centroid, full unit postcode or postcode sector with decreasing accuracy.
Improvements in geocoding and data quality are driving a renewed confidence in this geospatial intelligence, but while the result of the geocoding operation is a precisely located risk, whether that location is accurate or not is often overlooked. In the UK example to the left, one address has been geocoded to three levels of accuracy; all with the same coordinate precision, yet the location is different.
How does geocoding affect the analysis results?
The latest high-quality flood models allow the intensity of hazard and risk to property to be differentiated by individual building.
Where coordinates are provided, the catastrophe model takes these as the true location of the property and matches it to the nearest building footprint. Where this location is based on poor geocoding, this will give the incorrect result and the risk to that property will be incorrectly assessed. Overestimation of risk will give the underwriter an uncompetitive premium, while underestimating the risk could lead to a high-risk property being added to the portfolio.
Using our earlier example, the annual average cost (AAL) of flooding using each of the locations is shown below. In this case, if the premium is calculated based on the AAL for the postcode sector, this may be vastly underestimating the true cost flooding.
How do I ensure I have the most accurate view of risk?
Most geocoders return both a geocoding level and geocoding confidence with the coordinate result. Where a property is geocoded to street, postcode centroid or postcode sector level, these should be enriched with additional information and the geocoding process repeated, or discarded and analysis carried out at postcode resolution.
On average, postcodes in the UK contain 14 houses. While running at lower resolution sounds counter intuitive, it's better to let the catastrophe model sample the uncertainty in property location across all the houses within the postcode and provide an accurate, albeit generalised, estimation of risk rather than use incorrect data at coordinate level.
Where building centroid or individual street address geocoding is provided, good models will identify the correct building and calculate risk based only on the hazard and vulnerability appropriate to that property. This ensures insurers and reinsurers have the most competitive pricing and the most accurate view of both property and portfolio level risk.
This topic was presented at the ESRI Georisk conference in London on 13 March 2018. For more information, the slides are available online here.
If this has got you thinking about geocoding in your portfolio, read our post on building level analysis for building level exposure. Alternatively, get in touch to find out how our latest UK Flood Model can help you identify the most accurate view of your risk.
