An integrated regression analysis and time series model for construction tender price index forecasting

Abstract

Clients need to be informed in advance of their likely future financial commitments and cost implications as the design evolves. This requires the estimation of building cost based on historic cost data that is updated by a forecasted Tender Price Index (TPI), with the reliability of the estimates depending significantly on accurate projections being obtained of the TPI for the forthcoming quarters. In practice, the prediction of construction tender price index movement entails a judgemental projection of future market conditions, including inflation. Statistical techniques such as Regression Analysis (RA) and Time Series (TS) modelling provide a powerful means of improving predictive accuracy when used individually. An integrated RA‐TS model is developed and its predictive power compared with the individual RA or TS models. The accuracy of the RA‐TS model is shown to outperform the individual RA and TS models in both one and two‐period forecasts, with the integrated RA‐TS model accurately predicting (95% confidence level) one‐quarter forecasts for all the 34 holdout periods involved, with only one period not meeting the confidence limit for two‐quarter forecasts.