Casualty Actuarial Society

Research

ASTIN Bulletin, Volume 33, No.2 November 2003

Interest-Rate Changes And The Value Of A Non-Life Insurance Company
Thomas Albrecht, Allianz AG

This paper argues that interest rate changes can have a significant impact on the appraisal value of a non-life insurance company, even if assets and liabilities are duration-matched. The im-pacts are shown to be positive or negative, depending on the underlying parameters, including reserve adequacy, combined ratio, business growth rate, asset allocation, risk capital relative to premium income, and the correlation between interest rate and technical insurance results.

The full article is available on the CAS web site.

Common Poisson Shock Models: Applications to Insurance and Credit Risk Modeling
Filip Lindskog and Alexander McNeil

The authors present studies of common Poisson shock models (more extensive treatment than in the Wang' 1998 PCAS paper). A common shock is a happening which affects multiple risks, like a recession increasing the probability of bond defaults, or an earthquake hitting several lines of business as well as creating an asset hit. While such models are intuitively appealing, parameterizations can be challenging.

The full article is available on the CAS web site.

Asymptotic Dependence of Reinsurance Aggregate Claim Amounts
Ana Mata

The author investigates asymptotic dependence at the tails of bivariate claim frequency models (similar to a definition given by Gary Venter's PCAS paper - Tails of Copulas). Even though the linear correlation coefficient may be small, the dependence ratio converges to a non-trivial constant at the tails (corresponds to high retentions).

The full article is available on the CAS web site.

Chain Ladder Bias
Greg Taylor

Assumptions are discussed under which the chain-ladder method is biased or unbiased. When the losses for a cell in the triangle are a random effect plus a factor times the previous lag losses for the year, it is unbiased. In the case where the losses for the ith lag and jth accident year are independent variations from a row-column expectation figj the method is biased. This was the assumption used to generate losses in Stanard's famous reserve simulation test.

The full article is available on the CAS web site.

Favorable Estimators for Fitting Pareto Models: A Study Using Goodness-of-Fit Measures
with Actual Data

Vytaras Brazauskas and Robert Serfling

Fitting methods called generalized median and trimmed mean are applied to estimating Pareto parameters and found to be better than maximum likelihood according to several measures.

The full article is available on the CAS web site.

Setting a Bonus-Malus Scale in the Presence of Other Rating Factors:
Taylor's Work Revisited

Sandra Pitrebois, Michel Denuit and Jean-Fran├žois Walhin

Taylor developed a simulation approach to applying experience rating within an auto class plan. This paper shows how to do it analytically.

The full article is available on the CAS web site.

Shaun Wang and Gary Venter