Answers are grounded in this paper and cite their sources.
The distribution of insurance claims in a given time period is usually regarded as a random sum. This paper sets up a time series model for the value of the claims and combines it with a model for the number of claims. Thus past observations can be used to make predictions of future values of the random sum, and the overall model ensures that they are discounted appropriately. It is shown that explanatory variables can be introduced into the model, and how it can be extended to handle several groups. The general approach is based on the recently developed structural time series methodology.
No takes yet. Share an insight, caveat, or question.
Harvey et al. (1989) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: