In this paper, we first propose a statistical model, called the Coherent Incurred Paid model, to predict future claims, using simultaneously the information contained in incurred and paid claims. This model does not assume log-normality of the levels (or normality of the growth rates) and is semi-parametric since it only specifies the first and the second moments; however, in order to evaluate the impact of the normality assumption, we also propose a benchmark Gaussian version of our model. Correlations between growth rates of incurred and paid claims are allowed and the tail development period is estimated. We also provide methods for computing the Claim Development Results and their Values at Risk in the semi-parametric framework. Moreover, we show how to take into account the updating of the estimation in the computation of the Claim Development Results. An application highlights the practical importance of relaxing the normality assumption and of updating the estimation of the parameters.
