A model of building price forecasting accuracy
The purpose of this research was to derive a statistical model comprising the significant factors influencing the accuracy of a designer's price forecast and as an aid to providing a theoretical framework for further study. To this end data, comprising 181 building contract details, was collected from the Singapore office of an international firm of quantity surveyors over the period 1980 to 1991. Bivariate analysis showed a number of independent variables having significant effect on bias which was in general agreement with previous work in this domain. The research also identified a number of independent variables having significant effect on the consistency, or precision, of designers' building price forecasts. With information gleaned from bivariate results attempts were made to build a multivariate model which would explain a significant portion of the errors occurring in building price forecasts. The results of the models built were inconclusive because they failed to satisfy the assumptions inherent in ordinary least squares regression. The main failure in the models was in satisfying the assumption of homoscedasticity, that is, the conditional variances of the residuals are equal around the mean. Five recognised methodologies were applied to the data in attempts to remove heteroscedasticity but none were successful. A different approach to model building was then adopted and a tenable model was constructed which satisfied all of the regression assumptions and internal validity checks. The statistically significant model also revealed that the variable of Price Intensity was the sole underlying influence when tested against all other independentpage xiv variables in the data of this work and after partialling out the effect of all other independent variables. From this a Price Intensity theory of accuracy is developed and a further review of the previous work in this field suggests that this may be of universal application.