Models for choice of residential location and place of work are often based on cross-section data and the concepts of random utility. The theory leads to a class of stochastic models with well-known properties. The strength of these models is the ability to predict the relative distribution of quantities. However, there is a growing opinion that these models are fairly weak when it comes to long-term forecasting and the prediction of absolute levels. For that reason there has been an increasing attention towards time series, as a tool for predicting levels. In this paper we introduce a dynamic framework in which a set of static random utility models is embedded in a dynamic error-correction-model . Basically the ECM approach evolves from the observation that demand and supply quantities can be decomposed into different processes, witch in general adjustment differently over time regarding speed and level of aggregation. In the labour and housing market we decompose into processes at the long run, which is labour demand and housing supply, and processes at the semi-long run, which is labour supply and housing demand. Firms and the public sector control the longest run whereas the semi-long run comes from decisions taken by households or individuals. This observation leads to a natural ECM framework where the semi-long run adjusts to the long run. As a consequence we allow for the existence of temporary dis-equilibrium, which might be the consequence of bottlenecks at the labour market because of excess demand.