A neural network approach to inflation forecasting: recent evidence for the USA and UK

摘要

Macroeconomic policy-makers regard the manipulation of money supply as an appropriate tool for controlling inflation. This paper provides recent evidence torn the USA of the inflation forecasting potential of two alternative measures of broad money, those constructed torn a traditional simple sum formulation and the more sophisticated weighted index number measures known as Divisia monetary indices. In this context the modelling technique used, of neural networks, is a novel one. The results confirm that Divisia aggregates outperform their simple sum counterparts as inflation forecasting tools.