Criticizing Dr. Mark Spoerer's report: "How many former forced labor workers who may be entitled for compensation will still be alive in 2000?"

Spoerer's report is supporting the position of the German government. This report has to be criticized in some points concerning its statistical aspects.

1st Point of Criticism

Based on an analysis of the Reutlingen data taken from the program "Zwangsarbeiter während des zweiten Weltkrieges in Reutlingen - Erfassung und Auswertung der Ausländer-Meldekartei". Spoerer's distribution of the East European forced labor workers (the so-called "Ostarbeiter"), which is based on this data set, could not be reproduced. After a thorough analysis of the data and an additional verification of the data set, statisticians of the University of Konstanz obtained a different age distribution of Polish and Soviet-Russian forced labor workers.

2nd Point of Criticism

In his report Spoerer claims that medical progress during the previous decades was strong and particularly advantageous for elderly people. Hence, their life expectancy and their chances of survival were increased. In this context Spoerer refers to an article which appeared in a German actuary journal ("Herleitung der DAV-Sterbetafel für Rentenversicherungen" by Schmithals und Schütz (1995) in Blätter der Deutschen Gesellschaft für Versicherungsmathematik).
In general it should be stated that for the issue at hand it would be more advisable to refer to official national statistics. But, if one decides to make use of data from actuaries one should definitely employ mortality rates from East European countries instead of German data. The reason is that Spoerer's aforementioned statement is accurate for Germany and some other European countries. However, it is completely wrong concerning the relatively large number of citizens from successor countries of the former Soviet Union. People from these countries represent a considerable share of the former forced labor workers seeking compensation. Mortality statistics of international actuaries reveal that mortality rates in successor countries of the former Soviet Union rose steadily and evenly across age classes after 1989 (cf. MacDonald et al. (1998): "An international comparison of recent trends in population mortality." British Actuarial Journal). This can be inferred from Table 1 and 2.

Data from the United Nations that was used in the Konstanz projection reflect this trend, if such a trend exists, but usually not as strong as it is seen by insurance mathematicians.

Table 1: Mortality rates (per 1000) in countries of the former Soviet Union, 1989 and 1993, male.

  Estonia Latvia Lithuania Russian Federation Ukraine
Age 1989 1993 1989 1993 1989 1993 1989 1993 1989 1993
55 - 59 20.3 27.4 21.2 30.0 19.1 24.7 22.0 31.2 19.3 24.6
60 - 64 32.0 38.7 31.8 41.3 26.8 33.4 33.6 44.3 29.7 33.5
65 - 69 43.3 51.6 43.5 53.2 40.5 46.0 47.8 58.1 43.5 48.8
70 - 74 61.1 66.6 62.3 74.1 52.9 67.4 64.9 77.4 59.5 68.5
75 - 79 92.1 97.0 90.2 98.0 76.1 80.8 99.2 111.6 91.8 93.3
80 - 85 139.3 152.3 135.9 150.4 111.4 115.8 146.5 158.1 138.9 148.8
over 85 231.5 247.1 225.1 238.4 207.8 230.1 229.6 239.0 228.5 244.8

Source: MacDonald et. al. (1998), British Actuarial Journal

Table 2: Mortality rates (per 1000) in countries of the former Soviet Union, 1989 and 1993, female.

  Estonia Latvia Lithuania Russian Federation Ukraine
Age 1989 1993 1989 1993 1989 1993 1989 1993 1989 1993
55 - 59 8.6 9.1 8.4 10.0 7.3 8.6 8.2 10.8 7.7 9.4
60 - 64 12.7 14.4 13.0 14.6 10.9 12.3 13.2 16.5 12.4 13.8
65 - 69 21.3 21.8 20.6 22.3 19.4 19.3 22.9 25.4 21.4 22.9
70 - 74 35.4 37.1 33.9 36.6 30.7 35.3 35.0 41.3 34.6 40.6
75 - 79 59.6 60.0 58.5 61.5 51.7 49.5 62.2 69.1 62.0 63.0
80 - 85 103.6 108.0 99.7 106.0 86.2 97.6 104.0 112.8 103.5 115.1
over 85 199.5 221.3 192.6 206.5 177.9 197.7 195.6 213.8 199.0 215.2

Source: MacDonald et. al. (1998), British Actuarial Journal

3rd Point of Criticism

Life tables for the immediate post-war years and for the present margin of the time frame pose a significant problem.

If one disregards the aforementioned second point of criticism for the moment being and instead follows Spoerer's argumentation that medical progress in successor states of the former Soviet-Union was strong, one stumbles across Spoerer's conclusion that an underestimation of mortality in the immediate post-war years is outweighed by an over-estimation of mortality in the recent years. This conclusion is questionable. In fact, both periods of time show different degrees of sensitivity. This is due to the fact that the former forced labor workers form a population that experiences no increases, but merely decreases. This incurs that a particular cohort exhibits a different age at the two different points in time.

An example is in order for further clarification. Let us focus on the case of a man from Belarus who was born in 1925 and survived the war. We assume a one-year conditional probability of survival of 99.78% (due to the lack of appropriate data) for one of the immediate post-war years. That is, the probability of dying amounts to 0.22%. With no doubt this incurs that we underestimate the probability of dying in his cohort. If the actual probability of dying had been twice as high, we would have overestimated - ceteris paribus - the share of his cohort in the projection by 0.22%. Let us now focus on the same cohort in 1996. The one-year conditional probability of dying (of survival) now amounts to 7.28% (92.72%). If we employed the corresponding probabilities from 1991, which are 6.07% and 93.93%, respectively, we would underestimate - again ceteris paribus - the share of the cohort by 1.21%.

This effect becomes more pronounced, the higher the age of the particular cohort is chosen. This is the driving force behind the fundamental differences between Spoerer and Jeske. Despite the fact that the effect during the immediate post-war years also increases due to the increasing age of the cohorts, its impact on the results of the projection is small. This is due to the fact that the relatively older cohorts are also subject to a higher mortality rate in the medium period.

4th Point of Criticism

Spoerer's attempt to infer a common distribution of age and gender of the Ukrainian forced labor workers in 1945 from the marginal distribution of the variable age in 1997 and the marginal distribution of the variable gender in 1945 is condemned to fail from a statistical point of view. Spoerer's proceeding could only be implemented correctly if independence of the two variables age and gender was assumed. However, this assumption is clearly rejected by the Reutlingen data. Neither independence is suitable nor the gender proportions of 1945 is valid for the 1997 data. If the two variables were independent or the structure of dependence is invariant from 1945 until 1997, one merely would need to multiply the probabilities, respectively frequencies, in order to describe the original distribution. In order to emphasize this effect, a scenario calculation ("Planspiel") was implemented.

It is crucial to note that Spoerer's attempt of a verification of the numbers of the Ukrainian National Foundation "Understanding and Reconciliation" is improper. Indeed, it is difficult to comprehend Spoerer's distribution of Ukrainian forced labor workers in 1997, since these data are questionable at all.

If one made a projection for the lower boundary of the number of Ukrainian forced labor workers in 1945 based on Spoerer's distribution in 1997, one would obtain a number of Ukrainian (not Soviet!) forced labor workers surpassing 3 million. Nevertheless, one should bear in mind that the true number cannot be calculated for reasons stated above. These conclusions show the absurdity of the age distribution of Ukrainian forced labor workers in 1997.

5th Point of Criticism

An indication for inaccurate data supplied by East European Foundation Initiatives can be found in Spoerer's report in footnote 30 where he states that the numbers of Polish forced labor survivors in 1999 is based on data from 1992. Spoerer quotes an oral information from a representative of the Polish Foundation Initiative who argued that of the 497,300 persons registered in 1992, approximately 400,000 were still alive in 1997. Hence, the number of Polish civil forced labor workers alive in 1992 (340,000) was multiplied by a factor of 0.8 accordingly.

Since Spoerer takes these numbers to be accurate, he assumes an average annual mortality rate for the Polish forced labor workers who are still alive in 1992 of approximately

3 %

during the time period from 1992 to 1999.

Obviously, a population of elderly people, whose youngest members are more than 60 years of age, does imply a higher mortality rate than 3 percent. Either Polish forced labor workers in the Third Reich consisted merely of children and young persons or the estimated number of survivors will exceed its true value by far. Interestingly, the Polish Foundation Initiative contradicts itself, as its latest estimations for the number of survivors induces an older population.
Similarly, an annual mortality rate of 10 percent, which entered the recent discussion, is incompatible with the aforementioned numbers for the period 1992-1999. In the Konstanz model the annual mortality rate for Polish forced labor workers exceeds 6.5 percent for 1997 and 7.5 percent for 1999, respectively.
Copyright 2000. Dr. Roland Jeske