Extreme Value Analysis of Mortality at Age Beyond 105

We have applied the method using the extreme value theorem to analyze the mortality at the extreme ages (>105) in Belgium, France, Germany, and Switzerland for the birth cohorts from 1881 to 1898. After examining the homogeneity, we analyzed the data in two steps: first at group level and secondly at subsets level for the female datasets. An automated selection procedure based on the sequential goodness-of-fit tests combined with forward stopping rule has been used to select the appropriate thresholds for each dataset. We found all the datasets have a negative tail index which supports the existence of an ultimate age. We also noticed that the estimated ultimate age is very sensible to the selection of the threshold and the associated shape parameter. Lastly, the one-year force of mortality and the survival probability at extreme ages have been elaborated, which provides a practical view on the closing of the life table.