Time-series based office market models from urban economics (Rosen 1984, Wheaton / Di Pasquale 1996) describe the market movement for entire cities, but they do not consider local heterogeneity. Cross-sectional models like hedonic price modelling and the adaptation of the hedonic models for vacancy rates are difficult to couple with forecasting results. However, office market research needs spatio-temporal information in the way that forecasting and a detailed spatial resolution are integrated. The microsimulation is a methodological approach that fulfils both requirements. The paper draws especially on the concept of location choice and land-use simulations from urban economics (e.g. Waddell 2003). The case study shown in the paper is Stuttgart, a city in Southern Germany with 4.325 office buildings. About 7.000 office occupiers are complemented by synthetically generated missing values. Occupiers (starting population) and buildings are stored in GIS. An iteration consists of the following steps: First, each user decides by a Monte-Carlo-Simulation whether to stay, to leave or to move. The probability for a decision is influenced by building properties, user types and location variables that are estimated with a multinomial logit regression and a mover data sample. Second, all moving and immigrating occupiers choose between coincidentally offered available buildings (binary logistic regression for different user groups). The simulation is programmed in Visual Basic for Applications and offers the parallel simulation of multiple scenarios. The statistical evaluation of these scenarios results in information about the risk of vacancy in individual buildings and the absorption time of newly built office clusters.