Introduction Results

2. Data and methods

The empirical analyses in this paper are based on two different sets of data. Data published by the German Statistical Office [Statistisches Bundesamt 1980-1998] are used to examine mortality patterns across all ages and to investigate the contribution of different age groups to changes in life expectancy. The second data set used in this study is the Kannisto-Thatcher Oldest Old Database [Kannisto 1994], a database that was specifically designed to investigate mortality at the ages of 80 and above.

A decomposition method [Pollard 1988, p266, equation 3] is used to analyze the contribution of different age groups to changes in life expectancy at birth. Results obtained with this method are reasonably accurate as long as the increases in life expectancy are modest, as is the case with East and West Germany. The decomposition method is applied for the whole period under consideration, 1980-1996, as well as broken down into three parts: 1980-1987 (prior to reunification), 1987-1992 (the period of reunification), and 1992-1996 (the period after reunification). The results obtained from the analyses of contribution of different ages to changes of life expectancy at birth are shown in Figures 2a (females, East Germany), 2b (females, West Germany), 3a (males, East Germany), and 3b (males, West Germany).

The decomposition of change in life expectancy at birth is based on absolute changes in age-specific death rates, and absolute changes tend to be greater at older ages, where death rates are high. In addition to the decomposition analysis, we present average annual rates of mortality decline [e.g. Wilmoth and Horiuchi 1999, p486] by age and sex in order to investigate relative mortality decline. The rates of relative change in mortality are shown in Figures 4a (females, East Germany), 4b (females, West Germany), 5a (males, East Germany), and 5b (males, West Germany).

When it comes to mortality at the oldest ages, the Kannisto-Thatcher Oldest Old Database is presumably a better source of information than mortality estimates published by the German Statistical Office, because it relies on the method of extinct generations. This method [Vincent 1951] builds up the history of a cohort (its size by age and sex) by summing up data on deaths beginning with the oldest individual for a cohort which has died out. The cohort history was chosen because central death rates are most accurately calculated when a closed group of persons is followed up from one exact age to the next. It is known that accuracy becomes increasingly important when the numbers are small, as is the case at the oldest ages.

Specifically, the Kannisto-Thatcher Oldest Old Database includes data on deaths by sex and by single year of age combined with the year of birth. The data begin at age 80 and start with the year 1950 for most of the 31 countries that are presently included in the database. There is no upper limit for age for most of the countries. For some countries, however, the age limit is 99 years. Kannisto [Kannisto, 1994] studied the quality and reliability of German mortality data and concluded that the data on deaths were accurate for both East and West Germany. Information on cause of death is not included in the database. Cause-of-death data (particularly for old decedents) have many problems. Comparative analyses involving cause-of-death data for the period prior to German unification are even more problematic because the two German states used rather different coding procedures [Brückner 1993, Dinkel 1999].

In the present study we developed Lexis maps based on these data for ages 80 and over and for the period 1950-1996 (see Figures 6 and 7). Lexis maps can be used to graphically display mortality changes over time and separately for different ages. The maps presented here are based on mortality indicators whose layers differ by color. The colors indicate different death rates in Figures 6 and different death rate ratios in Figure 7.


Introduction Results

Old-Age Mortality in Germany prior to and after Reunification
Arjan Gjonça, Hilke Brockmann, Heiner Maier
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871