Claims Reserving Manual, vol.1: Section B: Data and forecasting

Document description

This section introduces some of the main building blocks for claims reserving. To begin with, there is the important idea of making a projection of past experience into the future. Since the future never takes the trouble to conform properly with the past, any projection whatsoever will be subject to error. One needs, therefore, to understand the principles which can lessen the likely degree of error, and so bring credibility to the work.

Apart from those principles which make for stability, there is the matter of the data themselves and the actual methods of forecasting. These are not intrinsically difficult matters, but there is a fair amount of detail to be mastered. On the data side, a number of different quantities can be used in the projections, or as supporting evidence — not only claim amounts, but such items also as claim numbers, premium income and loss ratios. They can often be displayed in different ways in the search for pattern and regularity, and the concept of the development table is particularly important here. Then there is the question of data validation, and of how the classification of the risk groupings is to be made.

On the forecasting side, there are some surprisingly simple methods available. It is straightforward, almost intuitively obvious, to look for the average or trend which is present in a sequence of figures. The really vital question to ask is whether the available evidence supports the continuation of such average or trend into future periods. Although far more elaborate types of projection can be devised, it is these simple foundations on which they rest, and which should therefore first be thoroughly understood.

Contents: The projection of past experience; Data groupings: principle of homogeneity; The claims development table; Data quantities; Simple breakdowns of the claims pattern; Data systems and validation; Forecasting: simple averages and trends; Mathematical trendlines