The taxation of property based on market values requires frequent appraisals for a large number of properties. In light of the recent property tax reform discussion in Germany, it has been argued that a valuebased tax therefore cannot be implemented at a reasonable cost. In several other countries, however, mass appraisal systems based on statistical methods are used for property tax assessments. In this paper, we show how this could in principle be done in Germany, using transactions data that local surveyor commissions are obliged to collect by law. We discuss the regression techniques for estimating land values from such data and illustrate them by applying them to data from Berlin, Germany. We find that the methods are capable of producing land value estimates that match up well with expert based assessments.