Non-linear forecasting of stock returns: Does volume help?

Abstract

The testing for and estimation of non-linear dynamics in equity returns is a growing area of empirical finance research. This paper extends this line of research by examining whether a hitherto unconsidered variable, namely volume, imparts non-linear dynamics within equity returns and whether it has forecasting power. A significant amount of evidence supports a negative relationship between volume and future returns, which in turn suggests that volume could act as a suitable threshold variable. The results presented here provide evidence of a logistic smooth-transition model for four international stock market returns, with lagged volume as the threshold. Further, this model provides better out-of-sample forecasts than a corresponding logistic smooth-transition autoregressive model, a simple AR model and a random walk model based on a trading rule. In addition, this model also provides better forecasting performance in three cases against alternate non-linear specifications. This provides evidence in favour of non-linear dynamics, in contrast with previous evidence, which had suggested the relative failure of non-linear models in forecasting exercises.