Adam Podlaha, the head of Aon Benfield’s Impact Forecasting division, is responsible for overseeing the collection of the data and its implementation in creating the latest catastrophe and other models. He explained some of the developments in his field in an interview at the Reinsurance Rendezvous.
“We are developing models in areas where there is good coverage,” he said; “because we’re trying to improve the models, and bring in things, which many of them don’t have.” An example includes a “custom damage function” for a specific company, one that perhaps “covers greenhouses or churches.”
Impact Forecasting is also working on “models for areas and territories, which have no coverage,” he said. “We just released an Arabian Peninsula earthquake model, for example, and a Chilean earthquake and tsunami model, which is quite novel because of the correlation between earthquakes and tsunamis.”
As far as gathering the necessary data to construct those models, Podlaha said he wished it were as easy in other parts of the world as it is in the U.S. and Europe. For more exotic locales he said “it’s really cooperation between our own research people, who look for the data that is available, but also its cooperation with local partners, with local universities, local commercial organizations.” For the Chilean earthquake model Impact Forecasting worked with the University of Valparaiso.
It employs the same data gathering process in other areas. “We look always for local partners,” Podlaha said. “It’s very important to get them, and to actually get involved with them” in collecting the data. He explained that it’s often a case of working with local governmental offices, such as hydrological agencies in Thailand for flood models.
Once they can be shown what the model is about, and how it will help them more clearly understand local conditions and risks, they are usually very happy to cooperate in the model’s construction. “But, he added; “it takes time to gain their trust and work out the details of the cooperation to make sure they know how they will benefit from the model.”
Podlaha cautioned that “a model is a tool; you can’t expect them to be perfect; however, what people expect from the model is that it will perform well with real events.” He cited last winter in Europe, when a number of severe storms caused significant losses. In that case “we used those storms to validate our model.”
Following the events Impact Forecasting received the data to do so. “The day after we got the wind speed data; we could work out the map of wind; we plug it into our model, and then we can give the market early estimates of the losses, and then we can compare the estimates in a few weeks’ time with the loss estimates from PERILS, an organization that collects data across Europe for wind.” In fact Podlaha said they were even able to forecast what the potential loss levels would be a day before the storm actually hit the affected areas.
While wind and floods are the principle focus of its activities, Impact Forecasting is also engaged in modeling other types of risks. Podlaha said they’re working on models for terrorist attacks, based on Lloyd’s RDS [Realistic Disaster Scenario] program. “They [Lloyd’s] have two in the U.S. – Rockefeller Center and the New York Stock Exchange. These are basic scenarios defined by Lloyd’s, which are in concentric circles, and they incorporate losses.
“What we have done is to take these scenarios to a different level. We still use the same basic definition of the scenario itself, but we employ fluid dynamics modeling. We actually create a 3-D blast of, for instance, a two ton bomb in Rockefeller Center, and we work out what the likely losses will be.” He described the process as being a “more up to date model” of the actual blast.
In the case above, the results showed less of an impact further from the blast site, as it took into account the shielding effects of buildings surrounding the blast site and other data. “We had all of the parameters in place,” Podlaha said. The blast scenarios also used the “same tools,” i.e. computer simulations, as are used by “official organizations in the U.S., the fire department and the New York police.”
It’s an interesting technique, as Podlaha said “it could be applied anywhere.” The next step is to expand beyond the Lloyd’s scenarios. “Clients are coming to us and saying that they would like to do this blast modeling for our areas ‘accumulations.’ It’s almost custom modeling for their portfolio.” Those clients can include both insurance companies and Lloyd’s syndicates, who are required to report using the RDS scenarios.
Impact forecasting is also analyzing the extent of “connectivity,” Podlaha explained. “It’s about basically giving us access to our clients, and to models or platforms, independently of each other.” For example, an insurance company that wants to underwrite a risk in Austria, “and we have a model for it, and you already have a system where you can do this underwriting, we can supply our model in a way that makes it easier to underwrite in your own system.”
The purpose is to “make sure that you can access different models independently of the platform.” The project involves working with PERILS, the Zurich-based firm that “has been supplying the European market with both loss data and exposure data. To make it easier we have reformatted the data into our ELEMENTS format. People can just unload it from the website an run it through their model.” www.insurancejournal.com/news/international/2014/09/11/340168.htm
This makes it easier to use as it “minimizes the work in formatting data and preparing it for analysis, and you spend more time on actually doing the analysis and analyzing the results.” The process means people are able to base decisions on actualities, rather than on the process of obtaining them.
Was this article valuable?
Here are more articles you may enjoy.