Case Estimate and IBNR Uncertainty
There is uncertainty surrounding case estimates on known claims and also on the level of unknown claims yet to be made (the Incurred But Not Yet Reported, or IBNYR claims; this is also sometimes known as “Pure IBNR”). When calculating reserve uncertainty, it is appropriate to consider the import of each of these factors. It is not necessarily the case that the uncertainty of the case estimates is the same as the uncertainty of the IBNYR.
One approach is to consider case estimates as being a fixed data item. To the extent that the case estimates actually turn out to be incorrect, this will be seen in the reserves as Incurred But Not Enough Reported (IBNER) claims. The IBNER and the IBNYR claims together form the overall Incurred But Not Reported (IBNR) reserve. As such, establishing the variability of the total IBNR reserve implicitly includes the inaccuracies of the case estimates.
This approach allows the use of the incurred data. Incurred data is generally used for best estimate reserving in preference to the paid data since it contains more information. However, since the incurred position can go down as well as up, using it in an uncertainty study requires the ability to model negative claim increments, which rules out some of the possible variability models.
If the paid data is modelled instead, it is the overall reserve (case estimates plus IBNR) that is considered variable. Since paid claims should, in most cases, never decrease, more possible variability models are available to use. However, the information contained within the case estimates is discarded using this approach.
It is also possible to model the variability of the case estimate IBNER and the IBNYR separately before combining them into an overall total reserve variability.
In an ideal world, the results of modelling paid claims and of modelling incurred claims would be identical. However, this is also true for deterministic estimations of the reserve. In practise, the extra information that the case estimates provide – both about the location of the reserve and its variability – tends to mean that different results are obtained depending on the dataset chosen. It is important therefore to consider carefully the underlying mechanisms in choosing both your data and your models.
What is Reserving Uncertainty and why is it Important?
Calculating Uncertainty
Combining Models
