A neural network approach to inflation forecasting: the case of Italy

In this paper, a Divisia monetary index measure of money is constructed for the Italian economy and its inflation forecasting potential is compared with that of its traditional simple sum counterpart. The powerful and flexible Artificial Intelligence technique of neural networks is used to allow a completely flexible mapping of the variables and a greater variety of functional form than is currently achievable using conventional econometric techniques. Results show that superior tracking of inflation is possible for networks that employ a Divisia M2 measure of money. During a period of high financial innovation in Italy Divisia outperforms simple sum at both the AL and M2 levels of monetary aggregation. This support for Divisia is entirely consistent with findings based on standard econometric techniques. Divisia monetary aggregates appear to offer advantages over their simple sum counterparts as macroeconomic indicators. Further, the combination of Divisia measures of money with the artificial neural network offers a promising starting point for improved models of inflation.