Methods to coordinate survey samples in the context of German business statistics

Measuring the economic activity of a country requires high-quality data of businesses. In the case of Germany, this is not only required at national level, but also at federal state level and for different economic sectors. Important sources for high-quality business data are the business register and, among others, also 14 business surveys which are conducted by the Federal Statistical Office of Germany. However, the quality requirements of the Federal Statistical Office are in contrast to the interests of the businesses themselves. For them, answering to a survey's questionnaire is an additional cost factor, also known as response burden. A high response burden should be avoided, since it can have a negative impact on the quality of the businesses' responses to the surveys. Therefore, sample coordination can be used as a method to control the distribution of response burden while securing high-quality data. When applying already existing business survey coordination systems, developed by different statistical institutes, legal and administrative standards of German official statistics have to be taken into account. These standards consider different sampling fractions, rotation fractions, periodicity, and stratification of the aforementioned 14 business surveys. Therefore, the aim of this doctoral thesis is to check the existing business survey coordination systems for their applicability in the context of German official statistics and, if necessary, to modify them accordingly. These modifications include the introduction of individual burden indicators which aim to take the individual perception of response burden into account. For this purpose, several synthetic data sets have been created to test the application of the modified versions of the different business survey coordination systems through Monte Carlo simulation studies. These data sets include a large panel data set, reflecting the landscape of businesses in Rhineland-Palatinate and three smaller, synthetic data sets. The latter have been created with the help of the R package BuSuCo which has been developed within the scope of this thesis. The above mentioned simulation studies are evaluated based on different measures for estimation quality as well as for the concentration and distribution of response burden.