Correspondence: Nauzer Balsara, College of Business, Northeastern Illinois University, 5500 North St Louis Avenue, Chicago, Illinois 60625, USA. E-mail: n-balsara@neiu.edu

¹is a professor of Finance at Northeastern Illinois University, Chicago. He obtained his PhD in Money and Financial Markets from Columbia University, New York, in 1986.

²is a graduate student in Accounting at the University of Illinois, Chicago. He graduated from the University of California, Los Angeles in 2006 with a Masters in Computer Science.

³is an assistant professor of Accounting at Mercer University, Atlanta. She obtained her PhD in Accounting from the University of Alabama in 2003.

Received 3 September 2008; Revised 3 September 2008.

Abstract

The variance ratio test suggests that we cannot reject the random walk null hypothesis for three major US stock market indexes between 1990 and 2007. Moreover, we find that the naïve forecasting model based on the random walk assumption frequently generates more accurate forecasts than those generated by the autoregressive integrated moving average forecasting model. Consistent with this finding, we find that the regular application of three commonly used technical trading rules (the moving average crossover rule, the channel breakout rule and the Bollinger band breakout rule) underperform the buy-and-hold strategy between 1990 and 2007. However, we observe significant positive returns on trades generated by the contrarian version of these three technical trading rules, even after considering a 0.5 per cent transaction costs on all trades.