Recent dramatic changes to US wind models prompted a torrent of accusations that the modelling industry had acted irresponsibly. Bermuda Re spoke with the big three modelling firms about the implications.
Catastrophe modelling firms have proved themselves a valuable and increasingly integral part of the re/insurance landscape, with the industry an important aid to re/insurers as they look to understand, rationalise and price existing and emerging risk. Modelling firms are key partners, providing analytical support to the underwriting process, but this year’s changes to RMS’s US wind model, outlined in the firm’s version 11, attracted considerable criticism, in what is a rare blow to the modelling firms.
The criticism levelled at RMS seems to have been due not so much to the changes that were introduced by the new models, but the dislocation between expected and modelled exposures. Expected, minor changes proved to be markedly more significant, with what many in the industry termed ‘Hurricane RMS’ prompting five to 15 percent rate increases at the US midyear wind renewals on the back of the increased exposures outlined by the new RMS model.
The implications of RMS 11
Addressing the new version 11 models, Ryan Ogaard, senior vice president, model solutions at RMS, defended the company’s latest model release, stating that the firm had released several models—most of which had been well received—but admitted that “in spite of consultation with the industry, more outreach and pre-release analysis than we, or anyone, had ever done before, there were still a few disconnects between the expectation of change and the actual impact on some individual portfolios”, adding that RMS has plans to “enhance our cooperation through industry discussion”. He said that the firm was introducing technology to help prepare the industry for future model changes, with the next full model version set to be released in 2013, along with a new platform to replace and build upon its existing platform, RiskLink.
RMS model changes have had evident implications for the other two leading modelling firms, AIR Worldwide and Eqecat, creating both challenges and opportunities for the pair. The challenge has been in differentiating their offering from RMS’s ‘disconnected’ approach; the opportunity has been in building potential market share on the back of RMS’s troubles. But as Uday Virkud, executive vice president at AIR Worldwide made clear, for its part, AIR Worldwide would not be making any dramatic changes in response to RMS 11. “Fundamentally, everyone agrees that the risk hasn’t changed—it is the same as it was understood last year—only one of the yardsticks has gone through a change.” He said that the science of inland risk has been available to everyone, with perhaps the only explanation for the dramatic model changes being the fact that “someone was behind the curve in digesting these data”.
Virkud said that suggestions that there is now a “greater understanding of inland risk” were in fact untrue and that such science had already been applied to previous AIR Worldwide models accompanied by far less “dramatic change”. Addressing the specifics of inland wind risk, Milan Simic, senior vice president and managing director at AIR Worldwide, said that it was a “phenomenon that was well understood and modelled”. He said that AIR Worldwide has been updating its models and assumptions year-on-year, with its most recent update resulting in “only a one percent increase in risk exposure”, adding that “it is hard to understand why people aren’t updating their models year-onyear. You have to ask yourself, why would you sit on the science?”.
"Companies need to be more prudent in their use of models and understanding their strenghts and limitations."
Bill Keogh, president of Eqecat, spoke in a similar vein about Eqecat’s response to RMS 11: “We’re fortunate that we did not need to respond. We have reviewed and updated our US hurricane model six times over the past five years and as a result our global client base was not exposed to sudden and dramatic changes.” He said that Hurricane Ike had not changed the company’s view on “filling rates or inland windfields”, with the firm’s summer update of its wind models showing “modest decreases in Florida”.
Both AIR Worldwide and Eqecat said that events had proved to be consistent with their view of risk, but Virkud warned against an over-reliance on model assumptions and a single model provider. “Companies need to be more prudent in their use of the models and understanding their strengths and limitations so that they don’t get so shocked when something changes like this,” he said. And he pointed the finger at those re/insurers that had been relying on only the one model as being the culprits behind the current lambasting of the modelling firms. “You can’t simply label the whole modelling industry as irresponsible because one firm has made a big change.”
Responding to views expressed by AIR Worldwide and Eqecat, RMS said that inland wind had not in fact been the main driver of change in model results, adding that “model change will not always be incremental. We’ve seen situations where real events have not followed model expectations. We feel that it is our responsibility to change our view of risk when we get new information, rather than wait until we see the wreckage in the morning”, said Ogaard.
Re/insurers have been developing proprietary catastrophe models for some time, with these internal systems providing them with further detail and the tools with which to examine and interpret risk and pricing. However, recent changes to RMS wind model exposures have encouraged some reinsurers to redouble their efforts to build such proprietary models. Asked whether this development represents a concern to the modelling industry, Ogaard said that RMS does not really see proprietary models as a threat. Rather he said that “RMS feels that we can help our clients understand their own internal models”, with the RMS platform being developed to “accommodate other models”. He added that alternative benchmarks are a valuable tool for understanding catastrophe risks and that “testing data and assumptions is healthy”.
Keogh takes a rather different view to internal models, arguing that while “it is important for capital providers to have their own view on risk”—drawn both from models and their own analysis—proprietary models will inevitably prove “hard work” for those building them. Instead, Keogh advocates accessing “multiple perspectives and exploring why models differ. That exploration is an education”.
Divergence, not convergence
With data capture and analysis and the science of predicting and interpreting risk improving year-on-year, it would seem safe to assume that modelling results across the big three firms would converge. However, talking with all three firms, it seems that this is not the expectation. As Virkud outlined “people are trying to simplify the questions to a very complex problem”. He said that models results can differ from county to county, state to state, across commercial and residential, making it imperative that re/insurers drill down into their assumptions.
Keogh, similarly, said that “the fundamentals of catastrophe risk are so fraught with uncertainty that I would be surprised if that happened”. And Ogaard echoed common sentiments arguing that firms have “differing levels of data and they will not all interpret the data the same way”. He added that modelling catastrophe risk is “so complicated that you can’t get to a single correct answer”. Rather the industry should accept that there will be “legitimate differences of opinion and that it is healthy to explore these differences”.
A panel approach proves unpopular
Following the changes to RMS’s US wind exposure ushered in by version 11, a number of re/insurers raised the prospect of turning to a panel of modelling firms in order to iron out potential future step changes. Addressing the possibility, it was evident that the modelling firms did not agree that such an approach would prove beneficial. As Keogh outlined: “Our role is to help clients set rational expectations about risk. ‘Groupthink’ is not the answer.” He added that if re/insurers were to “combine or blend results, it must be done very thoughtfully”. Ogaard likewise argued that “three black boxes are not necessarily better than one”. Instead, it is all about re/insurers’ ability to “open the black box and examine the critical assumptions within the model”.
Today, catastrophe models provide coverage and data on territories and risks around the world, with all three firms continuing to build new models in order to better understand and price emerging risk. New models are often built to satisfy demand from re/insurers, but increasingly modelling firms are taking the initiative in building new models and understanding. Turning to plans for expansion, all three firms indicated that they had further models in the pipeline. Simic detailed AIR Worldwide’s push to further develop its Caribbean wind model “to include practically all the Caribbean islands”; its work on ‘clustering’ in European windstorms and the development of its models into central and eastern Europe; and the expansion of its European earthquake model to include 30 countries. Virkud added that the firm now models risks in 90 countries around the world, taking in the full gamut of risk—“earthquake, wind, typhoon, hail, wildfires and terrorism”—with the firm continuing to build new models “as clients increase their interest in specific regions and perils”.
Keogh detailed the 95 countries and 181 models that Eqecat has in the market and, along with AIR Worldwide, singled out Asia as a region of particular interest for future development. Both firms highlighted Asian typhoon risk as a peril that had attracted significant interest following the Tohoku earthquake, with both AIR Worldwide and Eqecat developing models in response to events and calls from clients for models to better understand their exposures. In fact, in 2012 Eqecat will update 76 of their models, many of which are risk models in Asia, Keogh said. RMS, for its part, made improvements to its European wind and earthquake models. Ogaard said that the firm was also making a push on flood models—“a challenge that has not been fully addressed”— adding that RMS is not making a “huge push into new geographies, rather it is doing so into new perils and the exploitation of data”.
Strengthening their capabilities
Finally, turning to efforts to strengthen its capabilities, Ogaard said that the industry needs to increase the processing power of its models in order for them to be able to analyse data more quickly and provide a more condensed view of risk. “We need to able to interpret data more quickly, apply it more flexibly and manipulate the models,” he said. He added that models and their results needed to be available to all departments so that model results could be applied throughout an organisation. Keogh, for his part, highlighted Eqecat’s commitment to acquiring talent and developing software and models. He said that it was through ongoing collaboration and a “commitment to transparency” that the modelling industry could continue to provide invaluable support to its re/insurance clients.
cat modelling, RMS, AIR, Eqecat, US, reinsurance