Abstract
Typically, non-life insurance claims data is studied in claims development triangles which display the two time axes accident years and development years. Most stochastic claims reserving models assume independence between different accident years. Therefore, such models fail to model claims inflation appropriately, because claims inflation acts on all accident years simultaneously. We introduce a Bayes chain ladder reserving model which enables us to model claims inflation. In this model we derive analytical formulas for the posterior distribution, the claims reserves and their prediction uncertainty.
Similar content being viewed by others
References
Bühlmann H, Gisler A (2005) A course in credibility theory and its applications. Springer, Berlin
Clark DR (2006) Variance and covariance due to inflation. CAS Forum (fall), pp 61–95
Donnelly C, Wüthrich MV (2012) Bayesian prediction of disability insurance frequencies using economic factors. Ann Actuar Sci (to appear)
Gigante P, Picech L, Sigalotti L (2012) Claims reserving in the hierarchical generalised linear models framework (preprint)
Hertig J (1985) A statistical approach to the IBNR-reserves in marine insurance. Astin Bull 15(2):171–183
Jessen AH, Rietdorf N (2011) Diagonal effects in claims reserving. Scand Actuar J 1:21–37
de Jong P (2006) Forecasting runoff triangles. North Am Actuar J 10(2):28–38
Johnson RA, Wichern DW (1988) Applied multivariate statistical analysis, 2nd edn. Prentice Hall, Upper Saddle River
Kuang D, Nielsen B, Nielsen JP (2008) Identification of the age-period-cohort model and the extended chain-ladder model. Biometrika 95:979–986
Kuang D, Nielsen B, Nielsen JP (2008) Forecasting with the age-period-cohort model and the extended chain-ladder model. Biometrika 95:987–991
Kuang D, Nielsen B, Nielsen JP (2011) Forecasting in an extended chain-ladder-type model. J Risk Insur 78(2):345–359
Lee Y, Nelder JA, Pawitan P (2006) Unified analysis via H-likelihood. Generalized linear models with random effects. Chapman& Hall, Boca Raton
McNeil AJ, Frey R, Embrechts P (2005) Quantitative risk management: concepts, techniques. Princeton University Press, Princeton
Mack T (1993) Distribution-free calculation of the standard error of chain ladder reserve estimates. Astin Bull 23(2):213–225
Shi P, Basu S, Meyers GG (2012) A Bayesian log-normal model for multivariate loss reserving. North Am Actuar J 16(1):29–51
Venter GG (2007) Refining reserve runoff ranges. CAS Forum (summer)
Wüthrich MV (2010) Accounting year effects modeling in the stochastic chain ladder reserving method. North Am Actuar J 14(2):235–255
Wüthrich MV, Embrechts P, Tsanakas A (2011) Risk margin for a non-life insurance run-off. Stat Risk Model 28(4):299–317
Wüthrich MV, Merz M (2008) Stochastic claims reserving methods in insurance. Wiley, New York
Author information
Authors and Affiliations
Corresponding author
A. Appendix
A. Appendix
Proof of Theorem 2
For the proof of Theorem 2 we consider the moment generating function of \(\varvec{\xi }\). To do so, we choose \({\varvec{a}}=\left({{\varvec{a}}^{(1)}}^{\prime },\ldots ,{{\varvec{a}}^{(I+J)}}^{\prime }\right)^{\prime }\in \mathbb R ^{I(J+1)}\) with \({\varvec{a}}^{(k)}\in \mathbb R ^q\) for all \(k=1,\ldots ,I+J\). Then we obtain for the moment generating function of \(\varvec{\xi }\) (using the conditional independence of the \({\varvec{\xi }}^{(k)}\) in the fourth equation)
where \(\varvec{\varphi }\) and \(\Upsilon \) are defined in Theorem 2. Therefore, the unconditional distribution of \({\varvec{\xi }}\) is multivariate normal with parameters \(\varvec{\varphi }\) and \(\Upsilon \). This completes the proof.\(\square \)
Proof of Theorem 3
Formula (4.3) and Corollary 1 imply that
where the last equation is a direct consequence of log-normal distributions.\(\square \)
Proof of Theorem 4
By applying Corollary 1 and using the notation introduced in Theorem 3 we obtain for the covariance terms in formula (4.4)
By substituting \(\beta _{m,n}^{post}=\exp \left\{ {\varvec{e}}_{m,n}^{\prime }\varvec{\varphi }^{post}+\frac{1}{2}{\varvec{e}}_{m,n}^{\prime }\Upsilon ^{post}{\varvec{e}}_{m,n}\right\} \) we obtain for the conditional MSEP
In the last equation we multiply and divide each term \(C_{m,I-m}\) by \(\sum _{n=1}^{2^{J-(I-m)}}\beta _{m,n}^{post}\) and set, see Theorem 3,
Substituting \(\omega _{m,n}^{post}\) as defined in Theorem 4 provides the claim.\(\square \)
Proof of Corollary 2
For the coefficient of variation of the ultimate claim we obtain
The last step follows because \(\sum _{n=1}^{2^{J-\!(\!I-m\!)}}\!\omega _{m,n}^{post}\!=\!1\). The second upper bound for \({\rm Vco}\!\left(\!\left.C_J\!\right|\!{\mathcal D}_I\!\right)\) follows directly from
\(\square \)
Rights and permissions
About this article
Cite this article
Salzmann, R., Wüthrich, M.V. Modeling accounting year dependence in runoff triangles. Eur. Actuar. J. 2, 227–242 (2012). https://doi.org/10.1007/s13385-012-0055-3
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13385-012-0055-3