Testing the Waters - Argo Hosts Flood Model Event at Lloyd's

Testing the Waters

US flood risk has cascaded onto the insurance market’s agenda after four feet of water rained onto Texas in the wake of Hurricane Harvey.

High flood waters beneath a highway overpass

By Dr. Federico Waisman, SVP and head of analytics at Ariel Re, part of Argo Group

US flood risk has cascaded onto the insurance market’s agenda after four feet of water rained onto Texas in the wake of Hurricane Harvey.

In a market largely monopolised by the 50-year-old National Flood Insurance Program (NFIP), which covers around 5 million homes in the US, major reinsurers have had little depth of interest in US flood. That changed with the NFIP’s purchase of a $1 billion reinsurance layer last year after initially requesting support for a $7 billion programme.

Harvey’s cataclysmic rains have served to highlight the current big challenge for carriers with a mounting appetite for US flood risk: it is for the most part un-modelled. Read more about the challenges and solutions.

That will soon change. As it stands today there are four US flood models available in the market: AIR, property data specialists CoreLogic and AonBenfield’s Impact Forecasting have all launched models, alongside newcomer KatRisk of Berkeley, California, whose model will launch imminently. At present we know little about these models and their science and assumptions, but we can be certain that each will produce very different results.

Argo will put these four US inland flood models to the test on 2 November during a workshop event hosted by Lloyd’s. To begin, a technical representative from each modelling company will give a five-minute introduction to their product to explain, among other things, their research, the science behind the model and the data sources it incorporates. RMS was initially announced to participate in this event but the modelling firm withdrew as their model is still in production.

The excitement will continue with a side-by-side comparison. I will provide data for each modeller to use to demonstrate their new model, and several historical scenarios to run. When each has been put through its paces, the modellers will be given time to clarify or defend the results and assumptions. Finally, we will take questions from the floor.

The recent Hurricane Harvey will be one of the scenarios used to test against a hypothetical portfolio with locations across the US. Each model will deliver the modelled losses for this event and I will then compare the number of locations that each model shows as suffering a loss, the total quantum of losses, and exceedance probability curve for the region.

The goal of the exercise is to understand each model’s assumptions and science rather than to decide which is best, since each is unique and its efficacy will be based on its application. Some, for example, may run correlations with hurricane exposures, but others will not. Those who attend the workshop will gain valuable insights into a new line of business, which heretofore could not be modelled, and sufficient understanding to assess the four models, which will suddenly change that.

Watch the livestream of the event at argolimited.com.