Hostname: page-component-7b7567c8c-tnkh4 Total loading time: 0 Render date: 2026-11-16T08:30:40.862Z Has data issue: false hasContentIssue false

Experience rating with Poisson mixtures

Published online by Cambridge University Press:  16 July 2015

Garfield O. Brown
Affiliation:
Statistical Laboratory, Centre for Mathematical Sciences, Cambridge, CB3 0WB, UK
Winston S. Buckley*
Affiliation:
Mathematical Sciences, Bentley University, Waltham, MA 02452, USA
*
*Correspondence to: Winston S. Buckley, Mathematical Sciences, Bentley University, Waltham, MA 02452, USA. Tel: 781-891-2000; Fax: 781-891-2457; E-mail: winb365@hotmail.com
Rights & Permissions [Opens in a new window]

Abstract

We propose a Poisson mixture model for count data to determine the number of groups in a Group Life insurance portfolio consisting of claim numbers or deaths. We take a non-parametric Bayesian approach to modelling this mixture distribution using a Dirichlet process prior and use reversible jump Markov chain Monte Carlo to estimate the number of components in the mixture. Unlike Haastrup, we show that the assumption of identical heterogeneity for all groups may not hold as 88% of the posterior probability is assigned to models with two or three components, and 11% to models with four or five components, whereas models with one component are never visited. Our major contribution is showing how to account for both model uncertainty and parameter estimation within a single framework.

Information

Type
Papers
Copyright
© Institute and Faculty of Actuaries 2015 
Figure 0

Figure 1Plot of the number of observed claims for each group.

Figure 1

Figure 2Plot of the number of observed claims per unit exposure.

Figure 2

Figure 3Left: Model trace indicator. Right: Histogram of posterior model order. (a) Birth/death and split/merge model trace. (b) Birth/death and split/merge histogram. (c) Birth/death model trace. (d) Birth/death histogram. (e) Split/merge model trace. (f) Split/merge histogram.

Figure 3

Table 1Posterior model order.

Figure 4

Table 2Acceptance rates.

Figure 5

Figure 4Convergence diagnostics. KS, Kolmogorov–Smirnov.

Figure 6

Table 3Parameter estimates conditional on k=2.

Figure 7

Table 4Parameter estimates conditional on k=3.

Figure 8

Figure 5Probability (vertical axis) of data from group i (horizontal axis) being assigned to individual components conditional on k=2 (a) and k=3 (b).