Actuarial Expertise

ASTIN Working Party Releases Report on Reserving Practices for General Insurance Worldwide

Culminating work begun in spring 2015, the ASTIN Working Party on Non-Life Reserving Practices recently conducted and analyzed a survey on non-life reserving practices throughout the world.  The survey results were presented at the 2016 ASTIN Colloquium in Lisbon, Portugal, which was held May 31-June 3.. All those attending the colloquium were given a flash drive with working party results. A webinar on the working party results took place in September 2016.

Forty-two countries accounting for 87 percent of worldwide non-life premium participated in the survey. Countries were included from North America, Europe, Asia, Latin America, Oceania, the Middle East and Africa. The survey sought to understand key aspects of actuarial reserving practices, including what methods are used to provide the point estimate for reserves (referred to as deterministic methods), what methods are used to model reserve variability and what software is used in reserving.

Not surprisingly, the survey showed chain ladder as the most-used method for point estimates, followed by Bornhuetter-Ferguson. The loss ratio method also is quite widely used, and average cost and Cape Cod also enjoy wide use. Other emerging approaches and more statistically based methods, such as generalized linear models (GLMs), appear to be rarely used by actuaries for reserving at this time. These results are displayed in the “Main deterministic methods” graph,1 taken from the working party paper.

Another notable outcome was the low representation of U.S. companies in the survey. Only six U.S. companies participated, and those companies represent merely 20 percent of the U.S. market.

 

Bootstrap and Mack are the predominant methods for reserve variability. According to the survey, some countries favor Mack (the U.S. and Germany) while others prefer the bootstrap (Canada and Australia). The multivariate Merz and Wüthrich2 and GLM methods also are used, but MCMC (Markov Chain Monte Carlo3) appears to be used infrequently.

Based on the survey, the most frequently used tool in reserve analyses is Excel, though it is used less than 50 percent of the time. The next most common tool is specialized software, however, internally developed reserve specific software is also frequently used. A surprising conclusion for me is that R is not used at all in reserving among survey respondents, even though the chain ladder library provides many functions useful for reserving applications.

The majority of insurance companies in the survey perform reserve analyses quarterly, though some do them monthly. A small proportion, likely of smaller companies, conducted reserve analyses annually. Note that the survey respondents were predominantly working for medium ($50 M to $500 M of premium) and large (greater than $500 M of premium) companies.

Another notable outcome was the low representation of U.S. companies in the survey. Only six U.S. companies participated, and those companies represent merely 20 percent of the U.S. market. This compares to participation by 24 Canadian companies representing 80 percent of the market in Canada.

The working party’s survey results indicate an increase in the complexity of methods and technologies applied to reserving. For instance, there appears to be an increase in the use of reserve variability models. The working party believes that, in the future, more reserving methods will be applied at the individual claim level and the application of big data approaches will increase.

To access the report of the working party, visit the International Actuarial Association website at http://bit.ly/ASTINdocs.

This latest report is the second one released by an ASTIN Working Party. In 2015 the Big Data Working Party released its report on data analytics in non-life insurance. The work on data analytics continues, with the terms of reference (TOR) for a follow-up working party on predictive modeling expected to be released soon. The TOR describes the working party objectives, provides the working party’s schedule and is used to recruit working party volunteers.

Several years ago, ASTIN developed a goal to encourage the formation of working parties as a way to provide practical applied research to the international non-life actuarial community. Any member of ASTIN can organize a working party. The procedure for doing so is found at http://www.actuaries.org/ASTIN/Documents/ASTIN_WorkingParties_v6.pdf. A key component of initiating a working party is drafting the TOR. The working parties are one of the benefits of membership in ASTIN.

It’s anticipated that the work of one or more new working parties will be presented in 2017 at the ASTIN Colloquium in Panama City, Panama, August 20-24, 2017. More information on next year’s colloquium can be found at http://www.actuaries.org/panama2017/.  CAS members are encouraged to propose a presentation for the 2017 colloquium.


Louise Francis, FCAS, MAAA, is consulting principal for Francis Analytics & Actuarial Data Mining Inc. in Philadelphia.

1 http://www.actuaries.org/ASTIN/Documents/ASTIN_WP_NL_Reserving_Report1.0_2016-06-15.pdf

2 Merz and Wüthrich describe a multivariate approach for combining chain ladder and additive reserving methods in their 2009 paper “Prediction Error of the Multivariate Additive Loss Reserving Method for Dependent Lines of Business,” Variance 3:1.

3 See CAS Monograph No. 1, Stochastic Loss Reserving Using Bayesian MCMC Models, by Glenn Meyers. Myers has also published a number of Explorations columns in Actuarial Review describing the use of MCMC.