A Reexamination of Time-Varying Stock Return Predictability

Abstract

In predicting stock market returns, academic research has had its primary focus onmacroeconomic variables, and less attention has been paid towards technical indicators.The evidence of the stock return predictability is either absent or weak, and there arecases of contradicting evidence in the literature whether stock returns even are predictable.Over the last ten years, several papers find evidence that stock return predictability existsduring the bad economic states. These papers have used different approaches, whereasmost of them have been using NBER chronology of expansions and recessions, or investorsentiment index, to define good and bad economic times. Based on our knowledge, therehas been limited research regarding the use of bull and bear markets to determine thesemarket states. This thesis reexamines and extends previous studies on the time-varyingstock return predictability. Our research is similar to Huang et al. (2014), as we measurethe performance of different predictors by conducting a Newey-West-statistics derivedfrom one-state and two-state predictive regression for the in-sample forecast. However,our thesis is extended by using four different definitions of market states to examinewhether there is significant evidence of stock return predictability. Result of this thesispresents a mixed performance across the different macroeconomic variables and technicalindicators. Most of the predictors perform better in bull and bear markets compared toexpansions and recessions, and investor sentiment index. We have tried to compare ourresults to previous studies, but each study applied a combination of different datasets,approaches, and methodologies, and therefore, it would be impractical to compare thefindings.