Mixing Mortality Data Approach Using Czech Cohort Data

Very frequently, more sophisticated mortality models are required for modeling high age mortality where the lack of data is an important and significant problem. The lack of data not only corrupts mortality prediction attributes but may lead to wrong model estimates and inaccurate prediction results. In such situation, additional information of surrounding countries can be incorporated to the mortality model. Shorter prediction intervals as well as better prediction results can be obtained by using credibility mixing of mortality data. This paper presents comparison of two methods that can improve the estimates of old age mortality models. Cohort approach is used. The first improvement is recalculation of the population data with the Extinct cohorts method. Second improvement is incorporation of the data collected from surrounding countries using credibility theory. For all three datasets (original data + 2 improvements), extrapolation of the mortality rate is calculated assuming a parametric function for the increase with age. The estimation error is compared as well as the resulting expected residual lifetime and an annuity value.