The predictability of real office rents

The prediction and forecasting of office rents is undertaken routinely and formally by major surveying practices and consultants in the UK. These predictions and forecasts (which may be adjusted in-house) are used as tools in the investment decisions of the major institutions to inform on the relative performance of property market sectors/regions and the property market as a whole. This paper examines the predictability of the national Hillier Parker real office rent index by using a recursive modelling strategy and maximized selection criteria to choose a predictive model from a set of 15 in each year from 1985 to 1994. Predictions one year ahead are made using the chosen models. The results show that it is very difficult to choose the best predicting model in any one year. The selected models are often beaten by naive competitors such as 'no change' from, or 'same change' as, previous period and the ranking of a model in terms of its historic fit usually bears no relationship to its ranking in terms of how well it can predict relative to the other models in the set. Whilst the results are generally disappointing, in that the best predicting model is often not chosen, they indicate that there is some degree of predictability in the series.