Mortality at old ages: evaluation of data quality

Dmitri A. Jdanov, Max Planck Institute for Demographic Research
Domantas Jasilionis, Max Planck Institute for Demographic Research

The old age population in developed countries has increased very rapidly throughout the second half of the 20th century. Improvements in survival are pushing new limits. Despite this, internationally comparable high quality demographic data on old-age populations remain insufficient. Moreover, quality of mortality data for old ages is one of the most problematic issues. The Human Mortality Database (www.mortality.org) is recognized as one of the most reliable sources of data on old age mortality. Methods used for re-estimation of old age population in the HMD allow significantly improve quality of mortality estimates. However, estimates of old age mortality provided by the HMD may significantly differ from official statistics. There is still a room for further improvements of data quality. In our paper, we study old-age mortality patterns in the 40 countries that are currently included in the Human Mortality Database (HMD). We investigate four aspects of data quality: (1) quality and completeness of official data; (2) complexity of HMD methods required for specific data set to reconstruct detailed data and their influence on mortality estimates; (3) quality of HMD mortality estimates; and (4) comparability across time series and countries. As result, we introduced an objective classification of countries and periods simultaneously accounting for a whole set of selected data quality indicators. Our results suggest that the data on oldest-old mortality in industrialized countries is quite reliable. However, the data quality varies substantially across countries and time periods. Although the data quality has improved with time in the majority of the HMD countries, some countries (ex-USSR, Germany) show worsening data quality during the last 15-20 years. The presented paper is an extension of our previous work on evaluation of data quality in the Kannisto-Thatcher Database.