Structural time series analysis as a method for explorative real estate data analysis, simulation and forecasting

The use of time series analysis in real estate related data analysis and forecasting is not yet prevalent in Germany. Potential reasons are insufficient data, high fragmentation of the markets, and a practitioner-driven preference of qualitative forecasts. Deviations from the mainstream are dominated by standard linear regression techniques, which suffer from strong assumptions such as linearity and parameter constancy. A powerful alternative is offered by the Structural Time Series Analysis approach (STSA), that describes the components of time series in a clear and unambiguous manner and pursues a straightforward approach to forecasting. Furthermore it allows to explicitly include cyclical components, which are known to be a common feature of real estate time series. And while the mathematics involved in the estimation process are complex, the use of a software tool called STAMP (Structural Time Series Modeler and Predictor) renders this method applicable also for statistically trained practitioners. In this article a closer look is taken at the advantages and disadvantages in comparison to other dynamic modeling approaches, making use of a practical example with rent data of the Frankfurt office market.