Uncertainty

Calculating Uncertainty

There have been a number of techniques established for the calculation of reserve uncertainty.  EMB have pioneered the use of some of these – in particular the technique called bootstrapping.

Bootstrapping is a way of deriving an estimate of the full distribution of possible outcomes of an uncertain event using some observed data from that event.  It does this without the need for a predefined formal distribution, meaning that the technique is very flexible.

The key to being able to use a bootstrapping technique is to find a function of each data point such that the transformed data are independent and identically distributed. 

Once this has been achieved, new “pseudodata” are derived by resampling from the transformed data.  Any statistical measure can then be calculated from the pseudodata.  By repeating the process many thousands of times, one derives the distribution of the chosen statistical measure.

The technical challenges having been solved, this technique is straightforward to apply in practice.  Its use allows us to quickly identify important measures such as key percentiles of the reserve distribution.

Adaptations of the technique have allowed us to extend its approach to allow for additional information.  For example, we may wish to use the bootstrapping approach to define a distribution around an independently calculated best estimate reserve, perhaps established during a detailed quarterly reserving study.

What is Reserve Uncertainty and why is it Important?

Case Estimate and IBNR Uncertainty

Combining Models