General economic developments as well as recent fundamental changes in the Common Agricultural Policy will likely impact significantly on the European farm structure. Although a decline of total farm numbers continues to be the general observation, important differences occur across regions and farm types. These differentiated developments and their determinants are of high relevance for policy impact assessment at the regional level. The main objective of the analysis provided in this paper is to empirically identify whether regionally specific characteristics account for differences in regional farm structure development. This is exemplarily shown for German FADN regions. As methodological approach a combined time series, cross-sectional Markov chain analysis is applied. The non-stationary Markov model is estimated via generalised cross-entropy estimation technique with the transition probabilities being represented as multinomial logit functions of explanatory variables and their coefficients. Prior information on the transition probabilities is obtained from observed single farm movements of the FADN sample.