Predictive power of dividend yields and interest rates for stock returns in South Asia: Evidence from a bias-corrected estimator

Abstract

Predictive models of stock returns are often criticized for generating spurious predictability, unstable predictive relationship, and poor out-of-sample forecasting performance. This paper addresses these issues in the context of four major South Asian equity markets. We provide a bias-corrected estimate of the relationship of future stock returns to dividend yield and interest rate. We use a restricted vector autoregressive model, draw statistical inferences from a wild-bootstrap method with superior size and power properties, and allow model parameters to vary over time. Dividend yield is a significant predictor in both in- and out-of-sample (OOS) in two countries, while interest rate exhibits significant predictability in all four markets. Imposing theoretically motivated restrictions on model parameters appears to improve OOS predictability. Finally, time-variation in return predictability is found to be linked to countercyclical risk premium and persistence of the predictor variables.