The study examines four alternative rental forecasting models in the context of the London office market. We compare and contrast the forecasting ability of an ARIMA model, a Bayesian Vector Autoregression approach, a Structural model and a simultaneous equation model, in addition to some preliminary evidence with regard to combination models. The models are estimated using the CB Hillier Parker London Office index over the period 1977 to 1996, with out-of-sample testing undertaken on the following three years of data. Diagnostic testing is also conducted on the alternative models. The findings reveal that the Bayesian VAR model produces the best forecasts, while the ARIMA model fails to pick up on the large uptake in rental values during the testing period.