Bayesian Claims Reserving Methods In Non Life Insurance With Stan

This book first provides a review of various aspects of Bayesian statistics. It then investigates three types of claims reserving models in the Bayesian framework: chain ladder models, basis expansion models involving a tail factor, and multivariate copula models. For the Bayesian inferential methods, this book largely relies on Stan, a specialized software environment which applies Hamiltonian Monte Carlo method and variational Bayes.

This thesis investigates the usefulness of Bayesian modelling to claims reserving in general insurance. It can be divided into two parts: Bayesian methodology and Bayesian claims reserving methods. In the first part, we review Bayesian inference and computational methods. Several examples are provided to demonstrate key concepts. Deriving the predictive distribution and incorporating prior information are focused on as two important facets of Bayesian modelling for claims reserving. In the second part, we make the following contributions: 1. Propose a compound model as a stochastic version of the payments per claim incurred method. 2. Introduce the Bayesian basis expansion models and Hamiltonian Monte Carlo method to the claims reserving problem. 3. Use copulas to aggregate the doctor benefit and the hospital benefit in the WorkSafe Victoria scheme. All the Bayesian models proposed are first checked by applying them to simulated data. We estimate the liabilities of outstanding claims arising from the weekly benefit, the doctor benefit and the hospital benefit in the WorkSafe Victoria scheme. We compare our results with those from the PwC report. Except for several Markov chain Monte Carlo algorithms written for the purpose in R and WinBUGS, we largely rely on Stan, a specialized software environment which applies Hamiltonian Monte Carlo method and variational Bayes.

We revisit the gamma-gamma Bayesian chain-ladder (BCL) model for claims reserving in non-life insurance. This claims reserving model is usually used in an empirical Bayesian way using plug-in estimates for variance parameters, because this empirical Bayesian framework allows us for closed form solutions. The main purpose of this paper is to develop the full Bayesian case also considering prior distributions for variance parameters, and to study the resulting sensitivities.

In this monograph, authors Greg Taylor and Gráinne McGuire discuss generalized linear models (GLM) for loss reserving, beginning with strong emphasis on the chain ladder. The chain ladder is formulated in a GLM context, as is the statistical distribution of the loss reserve. This structure is then used to test the need for departure from the chain ladder model and to consider natural extensions of the chain ladder model that lend themselves to the GLM framework.

We propose a Bayesian spline model which uses a natural cubic B-spline basis with knots placed at every development period to estimate the unpaid claims. Analogous to the smoothing parameter in a smoothing spline, shrinkage priors are assumed for the coefficients of basis functions. The accident period effect is modeled as a random effect, which facilitate the prediction in a new accident period. For model inference, we use Stan to implement the no-U-turn sampler, an automatically tuned Hamiltonian Monte Carlo. The proposed model is applied to the workers' compensation insurance data in the United States. The lower triangle data is used to validate the model.

A wide range of topics give students a firm foundation in statistical and actuarial concepts and their applications.