Pooling the forecast outcomes from different models has been shown by Makridakis (1989), Clement (1989) and others to improve out-of-sample forecast test statistics beyond any of the individual component techniques. As well as conventional combining, a different approach to forecast combination is also followed in this paper viz. we use a method suggested more recently by Ridley (1997, 1999) in which negatively correlated forecasts are combined to see if this offers improved out-of-sample forecasting performance in property markets.