The role of green energy stock market in forecasting China's crude oil market: An application of IIS approach and sparse regression models

This study investigates the effectiveness of sparse regression models with their diverse specifications and the impulse indicator saturation (IIS) method in forecasting crude oil prices and their returns in China. The empirical findings showed that the IIS approach and sparse regression models are more effective than the benchmark of the no-change model. The IIS approach consistently outperformed the competing models and produced reliable results in terms of both ROS2 statistics and success ratio, as it is invariant to window size and time horizon. The results suggest that sparse regression models exhibit a notable decline in accuracy compared to the IIS approach over the long horizon. However, the selection of window size in the rolling window method plays a crucial role. Meanwhile, the sparse regression models are quite sensitive to window size; with a long horizon forecast and a small window size, the accuracy of the results is far less than with a broader window size. In contrast, the IIS method provides quite accurate forecasts regardless of window size and time horizon. The elastic smoothly clipped absolute deviation (E-SCAD) and elastic minimax concave penalty (E-MCP) remained more successful across the sparse models, possibly as a result of the high correlation among the set of predictors.