A natural estimate for the loss distribution is the observed (empirical) claim size distribution. However, if there have been changes in monetary values during the observation period, inflation corrected data should be used. For a sample of observations
the empirical distribution function (edf) is defined as:
![]() ![]() |
The empirical distribution function approach is appropriate only when there is a sufficiently large volume of claim data. This is rarely the case for the tail of the distribution, especially in situations where exceptionally large claims are possible. It is often advisable to divide the range of relevant values of claims into two parts, treating the claim sizes up to some limit on a discrete basis, while the tail is replaced by an analytical cdf.