Forecasting total factor productivity growth in the construction industry using neural network modelling

Total factor productivity (TFP) is a comprehensive industry-level productivity measure and determines an industry's competitiveness. This research proposes Jorgenson's method as an appropriate TFP measurement for the construction industry. It is then applied to estimate TFP growth in Singapore's construction industry. It is found that TFP growth tends to move in tandem with the construction business cycle. As a monitor of progress towards TFP growth, firstly factors affecting TFP growth of the construction industry of Singapore are identified and significant indicators are selected. Secondly, models using alternative techniques for forecasting TFP growth are developed and compared. As factors influencing TFP growth are complicatedly interacted, Artificial Neural networks (ANNs) are applied to solve such complex nonlinear mappings. To overcome overfitting caused by small dataset, Bayesian Neural Network (BNN) is adopted. The result shows that ANNs can model TFP growth more accurately than the regression technique. Finally, several policy implications are made.