Rating Firms Take Giant Leaps Forward On ERM, Capital Models

While industry consultants recently have suggested that the introduction of ERM criteria by rating agencies has spurred movement among insurers to use ERM processes across the p-c industry, S&P Managing Director Steven Dreyer is more modest about the role S&P has played.

"We have seen significantly increased appreciation for ERM in the p-c world, but I think it's driven more by companies' own bitter experiences" than by rating agencies, he said. The push came earlier for life companies, he said, noting that guarantees in universal life products "that came back to haunt them" prompted the companies to realize they needed a better understanding of their risks.

"With p-c companies, 9/11 was probably the first punch, and Katrina was the second," he said. "Despite the fact that companies had well-developed underwriting processes and investment portfolio analyses, [they] failed to recognize the correlation issues that became evident with 9/11."

Mr. Dreyer continued that with Hurricane Katrina, there were also "unexpected risks–flood and offshore energy risks–that weren't in any models."

A secondary driver has been regulation–more so in Europe than here, and more on the life insurance side, he said.

"To some extent, I suppose we have also been a catalyst. But we are cautious not to be in a position where we're driving companies" in any direction as to how "they need to manage themselves."

At Chicago-based Fitch Ratings, Keith Buckley, group managing director and global head of insurance, and Jeff Mohrenweiser, senior director, reveal grand visions of changing the ways that p-c insurers measure and manage their risks.

Fitch has developed a dynamic global economic capital model, known as Prism, which it will start using next year. After releasing extensive documentation for public comment in June, the model is now in a testing phase during which Fitch will allow insurers to input information and preview their results.

"Prism represents the first global stochastic economic capital model from any of the major credit ratings agencies in insurance," Mr. Buckley said.

Unlike static factor-based models, dynamic, or stochastic, models assess the ability of companies to withstand multiple stresses to underwriting and investment results simultaneously by running thousands of computer simulations of possible future environments.

Such dynamic models can capture the interaction of asset and liability risks, giving credits for diversification or increasing required capital to recognize risk correlations.

(Prism is not the first insurer capital model with dynamically modeled elements. In Sept. 2004, Moody's Investors Service developed a risk-adjusted capital model for p-c insurers that simulates exposures in four key areas–investments, reinsurance, reserves and underwriting. The Moody's model, using annual statement data as a basic data source, is used for U.S. p-c insurers, but Moody's indicated that the model could be adapted to non-U.S. data sets as available and appropriate.)

Mr. Buckley explained the motivations for developing Prism to NU recently. "If you think back to 1993, when the NAIC [National Association of Insurance Commissioners] produced its risk-based capital ratios, that really was a huge leap forward in how capital was measured from the perspective of a third-party observer."

He continued, "That was a while ago, and a lot of new technology and new theories have come out"–many tied in with stochastic modeling and dynamic financial analysis (DFA), he said.

Mr. Buckley noted that the dynamic model trend has picked up a lot of steam in Europe, where regulators are heading in that direction with an initiative known as Solvency II. "There, regulators are ultimately going to be asking companies to create their own economic capital models to demonstrate that balance sheets are in order," he said.

This, he said, created a real opportunity for Fitch "to lead the cause in coming up with a third-party observer capital model that shot ahead and enveloped all these trends. Our hope and our goal is that people view this as being as big a leap forward as the NAIC took with RBC."

Mr. Mohrenweiser said Fitch is now "leading the charge on DFA models. We expect this to give the p-c industry more impetus to start implementing those things."

This, he added, is sound actuarial financial modeling. "And if a rating agency is saying, 'This is what we're going to be basing your capital on,' then [insurers] have to start building their own models."

While Fitch analysts report that they have seen more U.S. p-c insurers starting to embrace stochastic models, "I don't see the p-c regulators moving in that direction," he said. "Even though, if you go on the Web site of the Casualty Actuarial Society, the first thing you see on there is DFA, DFA, DFA, it seems to fall on deaf ears with the regulators," Mr. Mohrenweiser said.

While Mr. Buckley said U.S. regulators may begin to feel pressure to change if Solvency II is well executed in Europe, he also believes Fitch's work could play a role here as well. "This is kind of utopia, but if we could show the NAIC that you can use a stochastic model, then that might inspire them," he said.

One notable feature of Prism is that while it uses local market data, the model measures capital adequacy across all sectors and countries on a consistent basis. That means that model outputs (required capital curves) will allow Fitch to compare different types of insurers–such as a U.S. auto insurer with a German life insurer.

Recognizing the increasing development of sophisticated models by insurers themselves, Fitch said its analysts will give credence to insurers' in-house model, where they exist, rather than using Prism in isolation. (In a second major June announcement, Fitch actually described a three-pronged approach, noting that Fitch will also consider regulatory capital requirements in evaluating capital adequacy.)

Fitch has also said that, going forward, beyond using Prism simply to evaluate the capital adequacy of insurers, it will form a part of Fitch's assessment of insurers' enterprise risk management. A report outlining Fitch's ERM methodology will be published in the third quarter.

Although S&P was first out of the chute with ERM, the rating agency has no plans to introduce its own dynamic simulation model. While S&P will stick with a static factor-based model instead, its capital model is nonetheless undergoing a massive overhaul, with changes to pricing risk and reserve volatility factors among those most likely to impact capital analyses for p-c insurers.

Explaining that the overall structure of the model will also change as S&P attempts to introduce more probabilistic-type analyses, Mr. Dreyer said, "We wanted to take a step toward the ultimate goal of a fully economic and stochastic capital model, which we believe can only be executed by the company–with the amount of detailed information that the company has."

Although this means insurers won't see diversification credits flow through the new S&P capital calculations, Managing Director Mark Puccia noted at S&P's annual conference in June that such credits may ultimately be available through the ERM analysis. If S&P judges a company to have "strong ERM," then it will undertake a review of any in-house economic capital model the company may be using, perhaps asking the company to stress certain assumptions a little bit more, he explained.

"If we are confident in the risk management process [and] the economic capital model, we'll rely on that more and more as a basis of our view of capital adequacy, as opposed to the static capital model we have in place," Mr. Puccia said.

Mr. Dreyer, the practice leader for North American insurance ratings, and Grace Osborne, a managing director at S&P, gave more information about the new capital model and the ERM initiative during an interview with NU late last month. Separately, in early June, representatives of Fitch described their latest changes. Excerpts of the interviews follow.

While the most striking difference between them is the static-dynamic divide in their capital models, the two firms delivered several common themes. For example, both explained that their newest models set them on a path to better communicate with rated companies and that the models now position them to be more futuristic and forward-looking in their review processes–anticipating possible risks rather than relying on data that described the past.

Importantly, both firms said their new capital models will not have any immediate impact on insurer ratings, with each describing phase-in periods that will allow for some give-and-take between rated companies and their analysts as both come to understand the results of the new models.