Recovering Localized Information On Agricultural Structure Underlying Data Confidentiality Regulations - Potentials Of Different Data Aggregation And Segregation Techniques

The modelling and information system RAUMIS is used for policy impact assessment to measure the impact of agriculture on the environment. The county level resolution often limits the analysis and a further disaggregation at the municipality level would reduce aggregation bias and improve the assessment. Although the necessary data exists in Germany, data protection rules (DPR) prohibit their direct use. With methods such as the Locally Weighted Averages (LWA), and with aggregation singling production activities into larger groups of activities, the data at the municipality level can be made publicly available. However, this reduces the information content and introduces an additional error. This paper’s aim is to investigate how much information is necessary to satisfactorily estimate Germany-wide production activity levels at the municipality level and whether the data requirements are still in compliance with the DPR. We apply Highest Posterior Density (HPD) estimation, which is easily able to include sample information as prior. We tested different prior information content at the municipality level. However, the goodness of the developed estimation approach can only be evaluated having knowledge about the population. Because the real population is not known to us, we took advantage of the special situation in Bavaria and derived a pseudo population for that region. This is used to draw information conforming to DPR for our estimation and to evaluate the resulting estimates. We found that the proposed approach is capable of adequately estimating most activities without violating the DPR. These findings allow us to extend the approach towards the Germany-wide municipality coverage in RAUMIS.