Technical trading rules and nonlinear time series models

Abstract

This study examines whether the two classes of nonlinear models, the self-exciting threshold autoregressive model (SETAR) and the autoregressive asymmetric moving average model (ARasMA), can explain the moving average rule results obtained from the data of the daily Toronto Stock Exchange 300 Index during January 1977-December 1995. First, this study estimates various models including the random walk and AR(l) as linear models and the SETAR and ARasMA as nonlinear models. Then, using the estimates of parameters and the fitted residuals, the bootstrap simulations are conducted to obtain simulated 'p -values' for the moving average rule results from the actual series under each null model. Our simulation results show that the SET AR model can explain the mean returns on the moving average rules reasonably well. However, the simulation results also show that none of our four null models can explain the standard deviations of returns on the moving average rules.