2. EXPLORATION OF THE NONRESPONDENTS IN THE EGLHS 
25 
The wave 1 sample started with cohort groups of equal size", due to the intentional cohort 
design of the life history study. In wave 2, five years later, the cohort composition changed 
slightly but not that much. 
Only 13% of the NRS persons belong to cohort 1930 whereas the starting percentage in wave 
1 was 25,4%. This large difference cannot be explained by natural mortality in the oldest 
group. Given this, an assumed mortality influence should also have occurred in wave 2 in 
which the interviews were performed only several months earlier. When one takes wave 1 as 
the basis for comparison, the percentage of cohort 1930 in the NRS file is only 50% of the 
expected proportion (13% NRS versus 25,4% wave 1), whereas the cohort reduction factor in 
wave 2 is only 91% (23,2% wave 2 versus 25,4% wave 1). I additionally checked this with 
official census data* When one takes the percentage for cohort 1930 of the microcensus 
1991 as the basis (=19,5%) and the corresponding percentage in 1995 from the Statistical 
Yearbook (=18,3%), we get a similar reduction factor of about 93,8%. 
To assess whether the different cohorts in the NRS file might represent different types of 
nonrespondents, I collected the reported reasons for loss of sample persons per cohort. 
Reasons for Loss in the Initial Sampling 
Firstly, we will look at the situation during the initial field. TABLE 2 below lists the 
percentages of loss by different reasons. 
This is of course not the representation of the German population pyramid. Due to the fact that I will not use 
the total initial sample as a whole for inferential statistics about the population, however, possible cohort 
stratification weights are neglected. The intention is to instead detect the differences or similarities of the NRS 
nonrespondents. 
* The same birth cohorts as in the EGLHS data were selected, i.e. 1930-1929-31, 1940-1939-41, 1950-1951- 
53, 1960-1961-63. The four groups altogether counted as 100%, so that the percentage e.g. 19,5% for cohort 
1930 indicates the relative proportion to the pool of the four cohort groups. Only the East German population 
was taken. It has to be mentioned that the census data counts the population by a regional concept. Therefore, 
East German population might contain West-East movers. In contrast to which, the life history study sampled 
using the concept of origin. This means that the sample only contains persons who were citizens of the GDR in 
1989. The sample does not, therefore, contain original West German citizens, although target persons might have 
moved to cities in West Germany. It is necessary to be aware of these conceptional differences so that margin 
distributions can only offer an approximate check. Sources for census information are: 
Mikrozensus 1991, 70% subsample (scientific public use file) of the complete census which is a 1% population 
sample. Neue Bundesländer, Berlin only the former Eastern part. 
Statistical Yearbook 1997, Federal Statistical Office Wiesbaden. Population data for 1995 on page 62.