Forecasting composite indicators with anticipated information: an application to the industrial production index

Abstract

Many economic and social phenomena are measured by composite indicators computed as weighted averages of a set of elementary time series. Often data are collected by means of large sample surveys, and processing takes a long time, whereas the values of some elementary component series may be available some time before the others, and may be used for forecasting the composite index. This problem is addressed within the framework of prediction theory for stochastic processes. A method is proposed for exploiting anticipated information in order to minimise the mean square forecast error, and for selecting the most useful elementary series. An application to the Italian general industrial production index is illustrated, which demonstrates that knowledge of anticipated values of some, or even just one, component series may reduce the forecast error considerably.