Predicting real estate rents: walking backwards into the future

Modelling, predicting and forecasting commercial rents are now seen as necessary and explicit processes in real estate investment. Decisions on the prospects for specific investments, the real estate portfolio and multi‐asset portfolio are made as a result of these processes and thus it is the accuracy of these models, predictions and forecasts in capturing future movements in rents that are implicitly tested in the marketplace. Despite the amount of theoretical and empirical research that has been conducted into modelling and predicting rents, it is unusual to find research which explicitly considers the predictive accuracy of models on an ex ante basis. This paper seeks to demonstrate the importance and possible value of such a procedure by examining the predictability of commercial rents in the office, industrial and retail markets of Great Britain over a real estate “cycle”. The paper concludes that theory appears to be a better indicator of the “correct” model structure than maximising historic fit. Often naïve competitors are better predictors than the model selection strategy employed.