V.1. Historical research
V.1.1. Sifting through the sources
As a first step, this means surveying the available contemporary records, documents and statistics (including estimates) about the extent of the dispossessions, and in similar format, the restitutions and compensations made after the end of the war.
At the same time, the primary sources, to the extent that they are still available, will have to be sifted in such a manner that an adequate strategy for handling the materials can be chosen: whether to survey all the data or use sampling procedures with subsequent extrapolation. From this, a specific project design will have to be developed in each case to permit a scientifically reliable answer to the questions posed.
Where the parent populations are very big, it is useful to resort to sampling. A sample may be very small if variance, in relation to the facts under investigation, is very small in the parent population.
Even a search of the total population would certainly not lead to an error-free result since the data themselves are likely to be defective and incomplete. It is true that one cannot state with certainty at what point the sampling error becomes negligible in relation to the errors in the sources. But it is clear that in many cases a point will soon be reached where more cases no longer mean a significant improvement of the result.
V.1.3. Completeness of archival sources as regards numbers of cases
The data which the Commission has to use are incomplete in many different respects, partly because some files were discarded or otherwise destroyed, and partly because some facts were not recorded at all. If, as in the case of the census of Jewish property in spring and summer 1938, a part of the population, viz. people with no property or property of less than 5,000 Reich marks, was left out of consideration, there remains no other way than to make up for the missing part by statistical methods. This can be done relatively easily and accurately in the case of the reporting of Jewish property by constructing a distribution curve, because the distribution was cut off at a certain point and moreover the number of missing cases is known.
If in the case of other questions the missing data are randomly distributed, an extrapolation can be made from the existing data to the whole aggregate.
The situation is difficult for the restitution files, where routine discarding of files frequently was neither random nor done on systematic criteria.
V.1.4. Completeness and reliability of the existing data in respect to individual components
The completeness of the data on assets, as far as they could be ascertained in 1938, is assured to different degrees. Immovables, assets of business enterprises, claims arising out of insurance contracts, bank deposits and other claims on banks (such as giro accounts, savings accounts and securities portfolios) as well as mortgage claims tend to be fairly complete.