Forecasting Taiwan stock return via crude oil price and asymmetrical volatility

In this study we aim to predict the excess returns of the Taiwan Stock Index (TAIEX) using crude oil prices and their asymmetric volatilities as predictors. The forecasting models employed are multiple linear regression models estimated by the ordinary least squares and exponentially weighted least squares methods. The sample period spans from January 2004 to December 2023. We evaluate model performance using metrics such as mean absolute error, root mean square error, out-of-sample R-squared, and cumulative annualized return. We summarize the empirical findings as follows. First, oil price volatility exerts asymmetric effects on subsequent monthly stock returns—positive during market booms and negative during recessions. In particular, bad crude oil volatility has a stronger impact on subsequent excess stock returns than good volatility. Second, increases in the interest rate, as well as in the growth rates of oil prices, exchange rate, and industrial production index, exhibit significant predictive power for stock returns. Finally, based on a buy–low–sell–high trading rule, we use the predicted stock price movement directions to compute cumulative annualized returns. Results show that most regression models outperform the TAIEX benchmark, suggesting that several models can offer valuable insights for investment decisions in the stock market.