What’s Up in the Rest of the World?
2008 ASTIN Colloquium Held in Manchester, England
By Louise Francis
Among CAS members, ASTIN’s colloquiua and publications have an image problem: they are perceived as highly abstruse and theoretical and of little practical value to North American actuaries. However, after attending the 2008 ASTIN Colloquium with its numerous timely, interesting, and practical presentations, I believe this reputation to be quite undeserved.
ASTIN (Actuarial STudies In Non-life insurance) is a section of the International Actuarial Association (IAA) dedicated to actuarial research, particularly in non-life (property and casualty) insurance. Climate change, extreme value modeling, and European Union reforms in solvency monitoring were just a few topics discussed when actuaries from around the world met in Manchester, England, in July for the 2008 ASTIN Colloquium. The world is shrinking and nothing brings that fact home more forcefully than the reality that climate change can affect the entire world—no place is immune.
Kicking off the Colloquium was a presentation on climate change by Julia Slingo, who is the director of the Centre for Global Atmospheric Modeling at the University of Reading in England. Her research strongly supports the hypothesis that we will experience significant global warming in future years, even under relatively optimistic scenarios. Ms. Slingo’s models indicate not only significant increases in temperature but also declines in the availability of food and water arising from the climate change.
On a related topic, Anthony Day discussed the concept of “peak oil,” the time when the world’s maximum rate of oil production (in the aggregate) is reached. After this point, oil production rates will decline. Mr. Day cited a number of authors who believe peak oil has already been reached. His paper presents the implications of a possible energy crisis, not only caused by insufficient oil and other sources of energy but due to the deterioration of the energy infrastructure. Mr. Day’s paper contains a number of very plausible scenarios.
One of the tools commonly used to model multivariate extreme values is known as a “copula.” Unfortunately much of the published literature on copulas is quite difficult to understand. In his plenary session, “Accounting for Extreme Value Dependence in Multivariate Data,” Professor Christian Genest of Canada’s Laval University provided one of the more pragmatic and user-friendly introductions to copulas for probability distribution tail modeling. Insurance data is typically heavy-tailed with dependencies for extreme (i.e., extremely large) values and failure to model such scenarios can have a significant impact on actuarial estimates and decisions based on them. Thus, a straightforward procedure for addressing the problem of modeling extreme values by ranking the data is of practical interest to actuaries. Genest covers how to test for extreme values and to parameterize the function used to model correlated extreme values. Had their modelers paid attention to extreme values, Long-Term Capital Management, the hedge fund notorious for almost causing an international financial crisis in the late 1990s, might still be around today, and the continuing subprime meltdown that began in 2007 might never have happened.
Just as in North America, there is considerable interest in quantifying variability in reserve estimates. Hui Liu and Richard Verrall presented their paper on using the bootstrap for quantifying loss reserve variability. Note that the bootstrap is one of the most popular procedures used in Europe to quantify the variability of loss reserves (England and Verrall’s 2001 paper describing the approach can be accessed from the CAS Web Site). Liu and Verrall modified the bootstrap procedure by incorporating the Munich chain ladder, a technique that European actuaries are apparently well acquainted with (introduced by Quarg and Mack in 2004).
Following Dr. Quarg’s presentation at the 2004 Casualty Loss Reserve Seminar (CLRS), the method Liu and Verrall espouse simultaneously uses both paid and incurred loss development triangles to estimate ultimate losses. The technique is so widely known outside North America that North American actuaries could benefit by becoming acquainted with it. The Liu and Verrall paper provides a concise summary of the key formulas; however, it does not present some of the motivation and insight into the approach contained in the original Quarg and Mack paper. Fortunately, the Quarg and Mack paper on the Munich Chain Ladder will be published in the next issue of Variance, so CAS members can acquaint themselves with the technique.
In another interesting session, Michael Fackler’s paper titled “Uncertainty of Past Inflation” pointed out that the (usually industry-wide) trend and inflation indices that actuaries regularly work with contain a large amount of uncertainty that is typically ignored in measuring variability. The paper demonstrates that the uncertainty from inflation can be quite great and should not be ignored. One surprising result of the paper is that, under certain circumstances, the inflation uncertainty is so large that older historic data should be downweighted in pricing estimates.
Capital allocation, the allocation of actual or notional surplus to segments of an insurance company’s portfolio, is an area of active research within the CAS—one with many competing methods and with little consensus on how to do it. Capital allocation is often a step in developing pricing estimates. Neil Bodoff contributes to the debate by presenting a novel approach to allocating capital in his paper, “Capital Allocation by Percentile Layer.” The approach is a percentile-based analysis of the contribution of each line (peril, policy, etc.) to the company’s value at risk (VaR) at multiple percentiles. In general, implementing the method would require a stochastic model such as dynamic financial analysis (DFA), which can model thousands of scenarios of company performance.
The model is relatively straightforward to understand and implement and is used to quantify contributions to incremental changes in the percentiles of the loss distribution. It recognizes that many lines/perils/policies/etc. contribute to the outcome at each percentile and that selecting only one percentile to base capital allocation on can be arbitrary or misleading.
Solvency II was another hot Colloquium topic. A plenary session and a number of papers discussed the actuarial impact of the impending changes to insurance companies and actuaries. Implementing Solvency II regulations will result in insurance companies utilizing DFA models. It should be noted that the International Account Standards Board and the Financial Accounting Standards Board have been developing standards of accounting—some significantly different from U.S. accounting—that all countries are to use in the near future. If, in fact, Canada already has solvency regulations that require a limited form of DFA.
Numerous other sessions featured papers (45 in all) on many topics, including catastrophe modeling, solvency modeling, dynamic financial analysis, VaR, capital allocation, and generalized linear models.
Social aspects of the Colloquium included a visit to the city of Chester and a tour of Tatton Park, an historic estate whose impressive mansion and magnificent gardens were recently featured in the film Brideshead Revisited. I also broadened my cultural horizons by visiting with some of the 170 delegates from the 30-plus countries represented. At the dinners and outings, I was generally among people whose first language was other than English, though nearly all of them spoke English well. While at the gala dinner at the Imperial War Museum North, for instance, I was able to swap stories with a Norwegian actuary about our fathers’ experiences during World War II. Visiting with delegates from so many different countries exposed me to points of view unlike those I’ve usually encountered at North American conferences.
I encourage CAS members to learn more about what’s happening in the rest of the actuarial world. Accepted papers from the 2008 ASTIN Colloquium can be found online.
Louise Francis is the consulting principal for Francis Analytics & Actuarial Data Mining Inc. in Philadelphia. The CAS Board elected her CAS Vice President-Research and Development. She begins her term this month.