¹Robert Dillard Teer Distinguished Professor of Business, East Carolina University, Greenville, NC 27858, USA Tel: +1 252 3286636; Fax: +1 252 328 6664; e-mail: stansells@mail.ecu.edu

Received 10 July 2003.

Abstract

In this study, neural network modelling procedures are used to forecast the direction of change in 19 sector stock indexes. Publicly available historical data on 19 economic indexes are used to perform the forecasts. Two separate sets of forecasts for each of the sector stock indexes are performed for the months January–May 2001 and for the months January–March 2003. The results are evaluated both on the basis of ability to forecast the direction of change in the stock index and on the basis of total gains in terms of index points. The results indicate that the models have some ability to forecast stock index changes.