We investigate the prediction of Italian industrial production. We first specify a model based on electricity consumption; we show that the cubic trend in such a model mostly captures the evolution over time of the electricity coefficient, which can be well approximated by a smooth transition model à la Terasvirta, with no gains in predictive power, though. We also analyze the performance of models based on data of different business surveys. According to basic statistics of forecasting accuracy, the linear energy-based model is not outperformed by any other single model, neither by a combination of forecasts. However, a more comprehensive set of evaluation criteria sheds light on the advantages of using the whole information available. Overall, the best forecasting performance is achieved by estimating a combined model which includes among regressors both energy consumption and survey data.
Published in 2000 in: Journal of Forecasting, v. 19, 5, pp. 419-440
